{
  "schema_version": "1.0",
  "snapshot_utc": "2026-10-27T18:06:24.511077+00:00",
  "scope": "Public task/experiment/condition aggregate results only; no global score. Original protocols and modes are separate.",
  "row_count": 1961,
  "model_labels": {
    "laya": "Laya English",
    "intern_decision_4b": "Intern-Decision-4B",
    "llama8b": "Llama-3.1-8B-Instruct",
    "minitaur8b": "Minitaur8B",
    "qwen35_4b": "Qwen3.5-4B",
    "gaussian_ts": "Gaussian Thompson (disclosed NIG reimplementation)",
    "jev_1_14": "Jev 1.14",
    "claude_opus_5_5": "Claude Opus 5.5",
    "claude_sonnet_5_5": "Claude Sonnet 5.5",
    "claude_haiku_5_5": "Claude Haiku 5.5"
  },
  "value_encoding": {
    "finite": "JSON number, exact original decimal text retained in value_raw",
    "defined_positive_infinity": "string Infinity, not clipped or converted to null",
    "undefined": "JSON null, original value_status and value_raw retained"
  },
  "rows": [
    {
      "row_id": "v00001",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier",
      "scope": "all",
      "target": "The observed ball was generated from the left urn; saved hidden event labels are used only for scoring.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.279465375775",
      "value": 0.279465375775,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00001",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/1/event_brier"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00002",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_additional",
      "scope": "all",
      "target": "true_generated_source_urn",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.279465375775",
      "value": 0.279465375775,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00002",
        "aggregate_source_id": "aggregate-source-0002",
        "source_field": "/full_code_default400/event_brier_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00003",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00003",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/1/invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00004",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0668048094958067",
      "value": 0.0668048094958067,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00004",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00005",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0668048094958067",
      "value": 0.0668048094958067,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00005",
        "aggregate_source_id": "aggregate-source-0003",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00006",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.06988993380218744",
      "value": 0.06988993380218744,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00006",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00007",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.06988943088799715",
      "value": 0.06988943088799715,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00007",
        "aggregate_source_id": "aggregate-source-0003",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00008",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "likelihood_logodds_weight",
      "metric_variant": "likelihood_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.04582489733927653",
      "value": 0.04582489733927653,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00008",
        "aggregate_source_id": "aggregate-source-0002",
        "source_field": "/full_code_default400/likelihood_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00009",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.29132085719423834",
      "value": 0.29132085719423834,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00009",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/1/posterior_mae"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00010",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae_full_unclipped_additional",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.29132085719423834",
      "value": 0.29132085719423834,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00010",
        "aggregate_source_id": "aggregate-source-0002",
        "source_field": "/full_code_default400/posterior_mae_full_unclipped_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00011",
      "task": "bayesian_urn_inference",
      "model": "laya",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "prior_logodds_weight",
      "metric_variant": "prior_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.12556742519668235",
      "value": 0.12556742519668235,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00011",
        "aggregate_source_id": "aggregate-source-0002",
        "source_field": "/full_code_default400/prior_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00012",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier",
      "scope": "all",
      "target": "The observed ball was generated from the left urn; saved hidden event labels are used only for scoring.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24129546530578133",
      "value": 0.24129546530578133,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00012",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/4/event_brier"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00013",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_additional",
      "scope": "all",
      "target": "true_generated_source_urn",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24129546530578133",
      "value": 0.24129546530578133,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00013",
        "aggregate_source_id": "aggregate-source-0004",
        "source_field": "/full_code_default400/event_brier_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00014",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00014",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/4/invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00015",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.22075133497128263",
      "value": 0.22075133497128263,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00015",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/4/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00016",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.22075133497128263",
      "value": 0.22075133497128263,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00016",
        "aggregate_source_id": "aggregate-source-0005",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00017",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.23823395285580773",
      "value": 0.23823395285580773,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00017",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/4/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00018",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.23823250300483778",
      "value": 0.23823250300483778,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00018",
        "aggregate_source_id": "aggregate-source-0005",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00019",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "likelihood_logodds_weight",
      "metric_variant": "likelihood_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.07022357629410228",
      "value": 0.07022357629410228,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00019",
        "aggregate_source_id": "aggregate-source-0004",
        "source_field": "/full_code_default400/likelihood_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00020",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.23451563502271347",
      "value": 0.23451563502271347,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00020",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/4/posterior_mae"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00021",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae_full_unclipped_additional",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.23451563502271347",
      "value": 0.23451563502271347,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00021",
        "aggregate_source_id": "aggregate-source-0004",
        "source_field": "/full_code_default400/posterior_mae_full_unclipped_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00022",
      "task": "bayesian_urn_inference",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "prior_logodds_weight",
      "metric_variant": "prior_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.3209490658824571",
      "value": 0.3209490658824571,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00022",
        "aggregate_source_id": "aggregate-source-0004",
        "source_field": "/full_code_default400/prior_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00023",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier",
      "scope": "all",
      "target": "The observed ball was generated from the left urn; saved hidden event labels are used only for scoring.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.23394649999999997",
      "value": 0.23394649999999997,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00023",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/2/event_brier"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00024",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_additional",
      "scope": "all",
      "target": "true_generated_source_urn",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.23394649999999997",
      "value": 0.23394649999999997,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00024",
        "aggregate_source_id": "aggregate-source-0006",
        "source_field": "/full_code_default400/event_brier_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00025",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00025",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/2/invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00026",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.13961907476186752",
      "value": 0.13961907476186752,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00026",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/2/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00027",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.13961907476186752",
      "value": 0.13961907476186752,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00027",
        "aggregate_source_id": "aggregate-source-0007",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00028",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.14434045334346593",
      "value": 0.14434045334346593,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00028",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/2/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00029",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.14433924574404955",
      "value": 0.14433924574404955,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00029",
        "aggregate_source_id": "aggregate-source-0007",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00030",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "likelihood_logodds_weight",
      "metric_variant": "likelihood_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.021990040689983704",
      "value": 0.021990040689983704,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00030",
        "aggregate_source_id": "aggregate-source-0006",
        "source_field": "/full_code_default400/likelihood_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00031",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24024652814285496",
      "value": 0.24024652814285496,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00031",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/2/posterior_mae"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00032",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae_full_unclipped_additional",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24024652814285496",
      "value": 0.24024652814285496,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00032",
        "aggregate_source_id": "aggregate-source-0006",
        "source_field": "/full_code_default400/posterior_mae_full_unclipped_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00033",
      "task": "bayesian_urn_inference",
      "model": "llama8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "prior_logodds_weight",
      "metric_variant": "prior_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.3941184611436034",
      "value": 0.3941184611436034,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00033",
        "aggregate_source_id": "aggregate-source-0006",
        "source_field": "/full_code_default400/prior_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00034",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier",
      "scope": "all",
      "target": "The observed ball was generated from the left urn; saved hidden event labels are used only for scoring.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.249375",
      "value": 0.249375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00034",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/3/event_brier"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00035",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_additional",
      "scope": "all",
      "target": "true_generated_source_urn",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.249375",
      "value": 0.249375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00035",
        "aggregate_source_id": "aggregate-source-0008",
        "source_field": "/full_code_default400/event_brier_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00036",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00036",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/3/invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00037",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.13540180074051023",
      "value": 0.13540180074051023,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00037",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/3/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00038",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.13540180074051023",
      "value": 0.13540180074051023,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00038",
        "aggregate_source_id": "aggregate-source-0009",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00039",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.13903152667917312",
      "value": 0.13903152667917312,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00039",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/3/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00040",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.1390295922756195",
      "value": 0.1390295922756195,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00040",
        "aggregate_source_id": "aggregate-source-0009",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00041",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "likelihood_logodds_weight",
      "metric_variant": "likelihood_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.009162866000365534",
      "value": 0.009162866000365534,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00041",
        "aggregate_source_id": "aggregate-source-0008",
        "source_field": "/full_code_default400/likelihood_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00042",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.2938360538345546",
      "value": 0.2938360538345546,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00042",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/3/posterior_mae"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00043",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae_full_unclipped_additional",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.2938360538345546",
      "value": 0.2938360538345546,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00043",
        "aggregate_source_id": "aggregate-source-0008",
        "source_field": "/full_code_default400/posterior_mae_full_unclipped_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00044",
      "task": "bayesian_urn_inference",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "prior_logodds_weight",
      "metric_variant": "prior_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.028384154645235977",
      "value": 0.028384154645235977,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00044",
        "aggregate_source_id": "aggregate-source-0008",
        "source_field": "/full_code_default400/prior_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00045",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier",
      "scope": "all",
      "target": "The observed ball was generated from the left urn; saved hidden event labels are used only for scoring.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.22987150000000003",
      "value": 0.22987150000000003,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00045",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/5/event_brier"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00046",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_additional",
      "scope": "all",
      "target": "true_generated_source_urn",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.22987150000000003",
      "value": 0.22987150000000003,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00046",
        "aggregate_source_id": "aggregate-source-0010",
        "source_field": "/full_code_default400/event_brier_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00047",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00047",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/5/invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00048",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24228924050112255",
      "value": 0.24228924050112255,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00048",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/5/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00049",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24228924050112255",
      "value": 0.24228924050112255,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00049",
        "aggregate_source_id": "aggregate-source-0011",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00050",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.26047561813320497",
      "value": 0.26047561813320497,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00050",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/5/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00051",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; timings are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.2604751540347934",
      "value": 0.2604751540347934,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00051",
        "aggregate_source_id": "aggregate-source-0011",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00052",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "likelihood_logodds_weight",
      "metric_variant": "likelihood_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.03546889273455109",
      "value": 0.03546889273455109,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00052",
        "aggregate_source_id": "aggregate-source-0010",
        "source_field": "/full_code_default400/likelihood_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00053",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24619969019819096",
      "value": 0.24619969019819096,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00053",
        "aggregate_source_id": "aggregate-source-0001",
        "source_field": "/models/5/posterior_mae"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00054",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae_full_unclipped_additional",
      "scope": "all",
      "target": "analytic_Bayes_posterior",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.24619969019819096",
      "value": 0.24619969019819096,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00054",
        "aggregate_source_id": "aggregate-source-0010",
        "source_field": "/full_code_default400/posterior_mae_full_unclipped_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00055",
      "task": "bayesian_urn_inference",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "generated_explicit_posterior",
      "metric": "prior_logodds_weight",
      "metric_variant": "prior_weight_original_no_intercept_OLS",
      "scope": "all",
      "target": "original_OLS_coefficient",
      "unit": "coefficient",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.6725302187255084",
      "value": 0.6725302187255084,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00055",
        "aggregate_source_id": "aggregate-source-0010",
        "source_field": "/full_code_default400/prior_weight_original_no_intercept_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00056",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.11225637181409295",
      "value": 0.11225637181409295,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00056",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00057",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.01391304347826087",
      "value": 0.01391304347826087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00057",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00058",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.16782608695652174",
      "value": 0.16782608695652174,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00058",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00059",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.1",
      "value": 0.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00059",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00060",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.1617391304347826",
      "value": 0.1617391304347826,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00060",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00061",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.043478260869565216",
      "value": 0.043478260869565216,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00061",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00062",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.1751207729468599",
      "value": 0.1751207729468599,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00062",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00063",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.11225637181409295",
      "value": 0.11225637181409295,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00063",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/24/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00064",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.9177048449510748",
      "value": 0.9177048449510748,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00064",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/24/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00065",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.0992503259452412",
      "value": 0.0992503259452412,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00065",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00066",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.03130434782608696",
      "value": 0.03130434782608696,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00066",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00067",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.1391304347826087",
      "value": 0.1391304347826087,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00067",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00068",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.024",
      "value": 0.024,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00068",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00069",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.136",
      "value": 0.136,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00069",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00070",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.11739130434782609",
      "value": 0.11739130434782609,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00070",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00071",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.12608695652173912",
      "value": 0.12608695652173912,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00071",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00072",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00072",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00073",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.144",
      "value": 0.144,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00073",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00074",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.026570048309178744",
      "value": 0.026570048309178744,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00074",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00075",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.17270531400966183",
      "value": 0.17270531400966183,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00075",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00076",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "2.342446202205846",
      "value": 2.342446202205846,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00076",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00077",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.368880121283347",
      "value": 2.368880121283347,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00077",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00078",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.290164678920686",
      "value": 2.290164678920686,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00078",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00079",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.384173007813883",
      "value": 2.384173007813883,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00079",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00080",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.3338087614966945",
      "value": 2.3338087614966945,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00080",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00081",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "2.3142687777096027",
      "value": 2.3142687777096027,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00081",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00082",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.3694338330836304",
      "value": 2.3694338330836304,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00082",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00083",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "2.4043110382429593",
      "value": 2.4043110382429593,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00083",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00084",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.2985470009691555",
      "value": 2.2985470009691555,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00084",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00085",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "2.3812791453143105",
      "value": 2.3812791453143105,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00085",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00086",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "2.309384991278646",
      "value": 2.309384991278646,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00086",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00087",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00087",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00088",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00088",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00089",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00089",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00090",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00090",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00091",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00091",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00092",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00092",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00093",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00093",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00094",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00094",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00095",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00095",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00096",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00096",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00097",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00097",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00098",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.7848413424799219",
      "value": 0.7848413424799219,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00098",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00099",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.8843453680165112",
      "value": 1.8843453680165112,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00099",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00100",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "2.8596805736636246",
      "value": 2.8596805736636246,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00100",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00101",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "3.92",
      "value": 3.92,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00101",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00102",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.205217391304348",
      "value": 2.205217391304348,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00102",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00103",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "4.064",
      "value": 4.064,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00103",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00104",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.272",
      "value": 2.272,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00104",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00105",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "2.9608695652173913",
      "value": 2.9608695652173913,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00105",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00106",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.1278260869565218",
      "value": 2.1278260869565218,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00106",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00107",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "2.62",
      "value": 2.62,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00107",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00108",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.332",
      "value": 2.332,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00108",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00109",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "4.269323671497585",
      "value": 4.269323671497585,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00109",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00110",
      "task": "multiple_cue_judgment",
      "model": "laya",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.8623188405797102",
      "value": 1.8623188405797102,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00110",
        "aggregate_source_id": "aggregate-source-0012",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00111",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.6557346326836582",
      "value": 0.6557346326836582,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00111",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00112",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.6365217391304347",
      "value": 0.6365217391304347,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00112",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00113",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.5965217391304348",
      "value": 0.5965217391304348,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00113",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00114",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.6608695652173913",
      "value": 0.6608695652173913,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00114",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00115",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.7565217391304347",
      "value": 0.7565217391304347,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00115",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00116",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.6630434782608695",
      "value": 0.6630434782608695,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00116",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00117",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.6159420289855072",
      "value": 0.6159420289855072,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00117",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00118",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.6557346326836582",
      "value": 0.6557346326836582,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00118",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/20/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00119",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.6733508245877061",
      "value": 0.6733508245877061,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00119",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00120",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.6513043478260869",
      "value": 0.6513043478260869,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00120",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00121",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.7260869565217392",
      "value": 0.7260869565217392,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00121",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00122",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.7391304347826086",
      "value": 0.7391304347826086,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00122",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00123",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.6582608695652173",
      "value": 0.6582608695652173,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00123",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00124",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.6026570048309179",
      "value": 0.6026570048309179,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00124",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00125",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.7041062801932367",
      "value": 0.7041062801932367,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00125",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00126",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.6733508245877061",
      "value": 0.6733508245877061,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00126",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/21/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00127",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.5253107931932697",
      "value": 0.5253107931932697,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00127",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/20/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00128",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.5763850451404061",
      "value": 0.5763850451404061,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00128",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/21/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00129",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.4783246414602347",
      "value": 0.4783246414602347,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00129",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00130",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.44782608695652176",
      "value": 0.44782608695652176,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00130",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00131",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.542608695652174",
      "value": 0.542608695652174,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00131",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00132",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.592",
      "value": 0.592,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00132",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00133",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.688",
      "value": 0.688,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00133",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00134",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.45217391304347826",
      "value": 0.45217391304347826,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00134",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00135",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.5643478260869565",
      "value": 0.5643478260869565,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00135",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00136",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0.6",
      "value": 0.6,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00136",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00137",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.74",
      "value": 0.74,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00137",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00138",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.24516908212560387",
      "value": 0.24516908212560387,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00138",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00139",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.36835748792270534",
      "value": 0.36835748792270534,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00139",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00140",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.46903520208604954",
      "value": 0.46903520208604954,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00140",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00141",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.4295652173913043",
      "value": 0.4295652173913043,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00141",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00142",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.5756521739130435",
      "value": 0.5756521739130435,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00142",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00143",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.468",
      "value": 0.468,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00143",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00144",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.672",
      "value": 0.672,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00144",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00145",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.44782608695652176",
      "value": 0.44782608695652176,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00145",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00146",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.5304347826086957",
      "value": 0.5304347826086957,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00146",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00147",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0.52",
      "value": 0.52,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00147",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00148",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.604",
      "value": 0.604,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00148",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00149",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.2391304347826087",
      "value": 0.2391304347826087,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00149",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00150",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.4214975845410628",
      "value": 0.4214975845410628,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00150",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00151",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.5695519909272746",
      "value": 1.5695519909272746,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00151",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00152",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.646654331765163",
      "value": 1.646654331765163,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00152",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00153",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.3874488310032793",
      "value": 1.3874488310032793,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00153",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00154",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.2266063834775125",
      "value": 1.2266063834775125,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00154",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00155",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.0585823347990482",
      "value": 1.0585823347990482,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00155",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00156",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "1.6956249350342287",
      "value": 1.6956249350342287,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00156",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00157",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.4153280756656879",
      "value": 1.4153280756656879,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00157",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00158",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "1.372534797212104",
      "value": 1.372534797212104,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00158",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00159",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.0141616166263787",
      "value": 1.0141616166263787,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00159",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00160",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "2.1465085730790294",
      "value": 2.1465085730790294,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00160",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00161",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.75502166060007",
      "value": 1.75502166060007,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00161",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00162",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.6243639730741246",
      "value": 1.6243639730741246,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00162",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00163",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.7153465093443374",
      "value": 1.7153465093443374,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00163",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00164",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.4449707735275328",
      "value": 1.4449707735275328,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00164",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00165",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.6804853329131269",
      "value": 1.6804853329131269,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00165",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00166",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.2432524436490486",
      "value": 1.2432524436490486,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00166",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00167",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "1.7006851207983182",
      "value": 1.7006851207983182,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00167",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00168",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.5087824909055256",
      "value": 1.5087824909055256,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00168",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00169",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "1.5550102314466054",
      "value": 1.5550102314466054,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00169",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00170",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.4121718879131286",
      "value": 1.4121718879131286,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00170",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00171",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "2.021637911727274",
      "value": 2.021637911727274,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00171",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00172",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.6555928774805202",
      "value": 1.6555928774805202,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00172",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00173",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00173",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00174",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00174",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00175",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00175",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00176",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00176",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00177",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00177",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00178",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00178",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00179",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00179",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00180",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00180",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00181",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00181",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00182",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00182",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00183",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00183",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00184",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00184",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00185",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00185",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00186",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00186",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00187",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00187",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00188",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00188",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00189",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00189",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00190",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00190",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00191",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00191",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00192",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00192",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00193",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00193",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00194",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00194",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00195",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.7843177104950882",
      "value": 0.7843177104950882,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00195",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00196",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.0480065000010654",
      "value": 1.0480065000010654,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00196",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00197",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.5440553220105357",
      "value": 1.5440553220105357,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00197",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00198",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "2.2711934139952064",
      "value": 2.2711934139952064,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00198",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00199",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.1274445893089962",
      "value": 1.1274445893089962,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00199",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00200",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.277391304347826",
      "value": 1.277391304347826,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00200",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00201",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.9321739130434783",
      "value": 0.9321739130434783,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00201",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00202",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.812",
      "value": 0.812,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00202",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00203",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.532",
      "value": 0.532,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00203",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00204",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "1.3304347826086957",
      "value": 1.3304347826086957,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00204",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00205",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.9704347826086956",
      "value": 0.9704347826086956,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00205",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00206",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "1.12",
      "value": 1.12,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00206",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00207",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.672",
      "value": 0.672,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00207",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00208",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.6763285024154588",
      "value": 1.6763285024154588,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00208",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00209",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.2161835748792271",
      "value": 1.2161835748792271,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00209",
        "aggregate_source_id": "aggregate-source-0014",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00210",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.1240221642764017",
      "value": 1.1240221642764017,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00210",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00211",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.2678260869565217",
      "value": 1.2678260869565217,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00211",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00212",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.8895652173913043",
      "value": 0.8895652173913043,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00212",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00213",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.196",
      "value": 1.196,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00213",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00214",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.644",
      "value": 0.644,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00214",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00215",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "1.308695652173913",
      "value": 1.308695652173913,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00215",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00216",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.991304347826087",
      "value": 0.991304347826087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00216",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00217",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "1.16",
      "value": 1.16,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00217",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00218",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.904",
      "value": 0.904,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00218",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00219",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.5748792270531402",
      "value": 1.5748792270531402,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00219",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00220",
      "task": "multiple_cue_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.1195652173913044",
      "value": 1.1195652173913044,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00220",
        "aggregate_source_id": "aggregate-source-0015",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00221",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.2571214392803598",
      "value": 0.2571214392803598,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00221",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00222",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.23391304347826086",
      "value": 0.23391304347826086,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00222",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00223",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.23478260869565218",
      "value": 0.23478260869565218,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00223",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00224",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.3739130434782609",
      "value": 0.3739130434782609,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00224",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00225",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.2252173913043478",
      "value": 0.2252173913043478,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00225",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00226",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.3502415458937198",
      "value": 0.3502415458937198,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00226",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00227",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.2391304347826087",
      "value": 0.2391304347826087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00227",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00228",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.2571214392803598",
      "value": 0.2571214392803598,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00228",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/23/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00229",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.9410467268670932",
      "value": 0.9410467268670932,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00229",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/23/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00230",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.22294654498044328",
      "value": 0.22294654498044328,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00230",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00231",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.19391304347826088",
      "value": 0.19391304347826088,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00231",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00232",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.23478260869565218",
      "value": 0.23478260869565218,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00232",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00233",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.236",
      "value": 0.236,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00233",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00234",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.112",
      "value": 0.112,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00234",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00235",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.3217391304347826",
      "value": 0.3217391304347826,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00235",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00236",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.22956521739130434",
      "value": 0.22956521739130434,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00236",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00237",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0.38",
      "value": 0.38,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00237",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00238",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.184",
      "value": 0.184,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00238",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00239",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.25",
      "value": 0.25,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00239",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00240",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.21497584541062803",
      "value": 0.21497584541062803,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00240",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00241",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "2.686880864933173",
      "value": 2.686880864933173,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00241",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00242",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.561888566033044",
      "value": 2.561888566033044,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00242",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00243",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.620954922062039",
      "value": 2.620954922062039,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00243",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00244",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.0394425616035323",
      "value": 2.0394425616035323,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00244",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00245",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.180807243358167",
      "value": 2.180807243358167,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00245",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00246",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "2.2358883654153634",
      "value": 2.2358883654153634,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00246",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00247",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "3.056654235622567",
      "value": 3.056654235622567,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00247",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00248",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "1.9420830382868532",
      "value": 1.9420830382868532,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00248",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00249",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.3868322124328376",
      "value": 2.3868322124328376,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00249",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00250",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "2.6465147491576198",
      "value": 2.6465147491576198,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00250",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00251",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "3.0879654738449664",
      "value": 3.0879654738449664,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00251",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00252",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00252",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00253",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00253",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00254",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00254",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00255",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00255",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00256",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00256",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00257",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00257",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00258",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00258",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00259",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00259",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00260",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00260",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00261",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00261",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00262",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00262",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00263",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.5324462729913648",
      "value": 1.5324462729913648,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00263",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00264",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "3.731406564998906",
      "value": 3.731406564998906,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00264",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00265",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.5415580182529336",
      "value": 1.5415580182529336,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00265",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00266",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.617391304347826",
      "value": 1.617391304347826,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00266",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00267",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.511304347826087",
      "value": 1.511304347826087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00267",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00268",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.492",
      "value": 1.492,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00268",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00269",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.656",
      "value": 1.656,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00269",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00270",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "1.3217391304347825",
      "value": 1.3217391304347825,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00270",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00271",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.4669565217391305",
      "value": 1.4669565217391305,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00271",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00272",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "1.22",
      "value": 1.22,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00272",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00273",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.64",
      "value": 1.64,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00273",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00274",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.3743961352657006",
      "value": 1.3743961352657006,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00274",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00275",
      "task": "multiple_cue_judgment",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.7801932367149758",
      "value": 1.7801932367149758,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00275",
        "aggregate_source_id": "aggregate-source-0016",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00276",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.3002248875562219",
      "value": 0.3002248875562219,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00276",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00277",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.17739130434782607",
      "value": 0.17739130434782607,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00277",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00278",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.3617391304347826",
      "value": 0.3617391304347826,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00278",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00279",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.26956521739130435",
      "value": 0.26956521739130435,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00279",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00280",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.3591304347826087",
      "value": 0.3591304347826087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00280",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00281",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.2246376811594203",
      "value": 0.2246376811594203,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00281",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00282",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.38768115942028986",
      "value": 0.38768115942028986,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00282",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00283",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.2832464146023468",
      "value": 0.2832464146023468,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00283",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00284",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.15391304347826087",
      "value": 0.15391304347826087,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00284",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00285",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.36434782608695654",
      "value": 0.36434782608695654,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00285",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00286",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.144",
      "value": 0.144,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00286",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00287",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.34",
      "value": 0.34,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00287",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00288",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.20434782608695654",
      "value": 0.20434782608695654,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00288",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00289",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.3356521739130435",
      "value": 0.3356521739130435,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00289",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00290",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0.16",
      "value": 0.16,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00290",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00291",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.3",
      "value": 0.3,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00291",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00292",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.25120772946859904",
      "value": 0.25120772946859904,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00292",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00293",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.358695652173913",
      "value": 0.358695652173913,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00293",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00294",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.9835823332846132",
      "value": 1.9835823332846132,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00294",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00295",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.0477408918830338",
      "value": 2.0477408918830338,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00295",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00296",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.930514944972667",
      "value": 1.930514944972667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00296",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00297",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.0171315549341053",
      "value": 2.0171315549341053,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00297",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00298",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.9768268684913348",
      "value": 1.9768268684913348,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00298",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00299",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "2.09000961870219",
      "value": 2.09000961870219,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00299",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00300",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.9463767286512526",
      "value": 1.9463767286512526,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00300",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00301",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "2.064066012492499",
      "value": 2.064066012492499,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00301",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00302",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.9680425636867647",
      "value": 1.9680425636867647,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00302",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00303",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "2.037229716021191",
      "value": 2.037229716021191,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00303",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00304",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.9283837948450044",
      "value": 1.9283837948450044,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00304",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00305",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00305",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00306",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00306",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00307",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00307",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00308",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00308",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00309",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00309",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00310",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00310",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00311",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00311",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00312",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00312",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00313",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00313",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00314",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00314",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00315",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00315",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00316",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.98917719401652",
      "value": 1.98917719401652,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00316",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00317",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "4.511550285969861",
      "value": 4.511550285969861,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00317",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00318",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.5102672750977835",
      "value": 1.5102672750977835,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00318",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00319",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.751304347826087",
      "value": 1.751304347826087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00319",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00320",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.3043478260869565",
      "value": 1.3043478260869565,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00320",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00321",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.74",
      "value": 1.74,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00321",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00322",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.38",
      "value": 1.38,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00322",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00323",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "1.9304347826086956",
      "value": 1.9304347826086956,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00323",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00324",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "1.3313043478260869",
      "value": 1.3313043478260869,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00324",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00325",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "2.98",
      "value": 2.98,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00325",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00326",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.568",
      "value": 1.568,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00326",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00327",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.7342995169082125",
      "value": 1.7342995169082125,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00327",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00328",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "1.2330917874396135",
      "value": 1.2330917874396135,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00328",
        "aggregate_source_id": "aggregate-source-0017",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00329",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.1555472263868066",
      "value": 0.1555472263868066,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00329",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00330",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.06",
      "value": 0.06,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00330",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00331",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.24695652173913044",
      "value": 0.24695652173913044,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00331",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00332",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.1391304347826087",
      "value": 0.1391304347826087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00332",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00333",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.20347826086956522",
      "value": 0.20347826086956522,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00333",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00334",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.06521739130434782",
      "value": 0.06521739130434782,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00334",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00335",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.18961352657004832",
      "value": 0.18961352657004832,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00335",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00336",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.1555472263868066",
      "value": 0.1555472263868066,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00336",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/22/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00337",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "1.058590894701496",
      "value": 1.058590894701496,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00337",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/22/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00338",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.1681877444589309",
      "value": 0.1681877444589309,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00338",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00339",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.07739130434782608",
      "value": 0.07739130434782608,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00339",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00340",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.2426086956521739",
      "value": 0.2426086956521739,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00340",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00341",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.152",
      "value": 0.152,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00341",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00342",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.332",
      "value": 0.332,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00342",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00343",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0.12608695652173912",
      "value": 0.12608695652173912,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00343",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00344",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0.20956521739130435",
      "value": 0.20956521739130435,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00344",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00345",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0.42",
      "value": 0.42,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00345",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00346",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0.468",
      "value": 0.468,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00346",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00347",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.026570048309178744",
      "value": 0.026570048309178744,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00347",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00348",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0.13647342995169082",
      "value": 0.13647342995169082,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00348",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00349",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "3.0120727126354185",
      "value": 3.0120727126354185,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00349",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00350",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "3.2200129582694377",
      "value": 3.2200129582694377,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00350",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00351",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.883338924711438",
      "value": 2.883338924711438,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00351",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00352",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.556052834923772",
      "value": 2.556052834923772,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00352",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00353",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.644653171899932",
      "value": 2.644653171899932,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00353",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00354",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "2.9244332378999656",
      "value": 2.9244332378999656,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00354",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00355",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.815952984481178",
      "value": 2.815952984481178,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00355",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00356",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "2.5195568522813914",
      "value": 2.5195568522813914,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00356",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00357",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.7988148736128948",
      "value": 1.7988148736128948,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00357",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00358",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "3.5282082311074245",
      "value": 3.5282082311074245,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00358",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00359",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "3.327347155014986",
      "value": 3.327347155014986,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00359",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00360",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00360",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00361",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00361",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00362",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00362",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00363",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00363",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00364",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00364",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00365",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00365",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00366",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00366",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00367",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00367",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00368",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00368",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00369",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00369",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00370",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00370",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00371",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "0.7712029909889679",
      "value": 0.7712029909889679,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00371",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00372",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "1.5417587559786625",
      "value": 1.5417587559786625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00372",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00373",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "3.298891786179922",
      "value": 3.298891786179922,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00373",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00374",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "3.84",
      "value": 3.84,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00374",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00375",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.8260869565217392",
      "value": 2.8260869565217392,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00375",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00376",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "3.42",
      "value": 3.42,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00376",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00377",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp1.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "2.492",
      "value": 2.492,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00377",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp1.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00378",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "230",
      "n": 230,
      "value_raw": "3.608695652173913",
      "value": 3.608695652173913,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00378",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00379",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1150",
      "n": 1150,
      "value_raw": "2.7295652173913045",
      "value": 2.7295652173913045,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00379",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00380",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "50",
      "n": 50,
      "value_raw": "2.48",
      "value": 2.48,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00380",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00381",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp2.csv:no_feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "250",
      "n": 250,
      "value_raw": "1.74",
      "value": 1.74,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00381",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp2.csv:no_feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00382",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_numeric",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "4.22463768115942",
      "value": 4.22463768115942,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00382",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_numeric/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00383",
      "task": "multiple_cue_judgment",
      "model": "qwen35_4b",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "collsiöö2023MCPL/exp3.csv:feedback/caldionine_level_verbal",
      "target": "original_human_numeric_response",
      "unit": "original_concentration_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "828",
      "n": 828,
      "value_raw": "3.710144927536232",
      "value": 3.710144927536232,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00383",
        "aggregate_source_id": "aggregate-source-0018",
        "source_field": "/by_experiment_phase_response_kind/collsiöö2023MCPL~1exp3.csv:feedback~1caldionine_level_verbal/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00384",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.4947916666666667",
      "value": 0.4947916666666667,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00384",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00385",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.4947916666666667",
      "value": 0.4947916666666667,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00385",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00386",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.7297495561754554",
      "value": 0.7297495561754554,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00386",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00387",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.7297495561754554",
      "value": 0.7297495561754554,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00387",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00388",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00388",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00389",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00389",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00390",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6204671774758026",
      "value": 0.6204671774758026,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00390",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00391",
      "task": "multi_attribute_choice",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "1.1365036999923177",
      "value": 1.1365036999923177,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00391",
        "aggregate_source_id": "aggregate-source-0019",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00392",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6753472222222222",
      "value": 0.6753472222222222,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00392",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00393",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6753472222222222",
      "value": 0.6753472222222222,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00393",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00394",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.5456727505555088",
      "value": 0.5456727505555088,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00394",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00395",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.5456727505555088",
      "value": 0.5456727505555088,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00395",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00396",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00396",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00397",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00397",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00398",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6680122394755017",
      "value": 0.6680122394755017,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00398",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00399",
      "task": "multi_attribute_choice",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "1.077778167964425",
      "value": 1.077778167964425,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00399",
        "aggregate_source_id": "aggregate-source-0020",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00400",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.5416666666666666",
      "value": 0.5416666666666666,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00400",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00401",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.5416666666666666",
      "value": 0.5416666666666666,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00401",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00402",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.7856294671769466",
      "value": 0.7856294671769466,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00402",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00403",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.7856294671769466",
      "value": 0.7856294671769466,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00403",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00404",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00404",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00405",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00405",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00406",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.7085091469925828",
      "value": 0.7085091469925828,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00406",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00407",
      "task": "multi_attribute_choice",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "1.1164620500057936",
      "value": 1.1164620500057936,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00407",
        "aggregate_source_id": "aggregate-source-0021",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00408",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.5503472222222222",
      "value": 0.5503472222222222,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00408",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00409",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.5503472222222222",
      "value": 0.5503472222222222,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00409",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00410",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.7220688128806454",
      "value": 0.7220688128806454,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00410",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00411",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.7220688128806454",
      "value": 0.7220688128806454,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00411",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00412",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00412",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00413",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00413",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00414",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6956827974936459",
      "value": 0.6956827974936459,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00414",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00415",
      "task": "multi_attribute_choice",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "1.102258056984283",
      "value": 1.102258056984283,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00415",
        "aggregate_source_id": "aggregate-source-0022",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00416",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.65625",
      "value": 0.65625,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00416",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00417",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.65625",
      "value": 0.65625,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00417",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00418",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6126204280717138",
      "value": 0.6126204280717138,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00418",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00419",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6126204280717138",
      "value": 0.6126204280717138,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00419",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00420",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00420",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00421",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "hilbig2014generalized/exp1.csv:choice/product_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00421",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/by_experiment_phase_response_kind/hilbig2014generalized~1exp1.csv:choice~1product_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00422",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.6849252019892447",
      "value": 0.6849252019892447,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00422",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00423",
      "task": "multi_attribute_choice",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "1.1103058700100519",
      "value": 1.1103058700100519,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00423",
        "aggregate_source_id": "aggregate-source-0023",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00424",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "4.049976210409848",
      "value": 4.049976210409848,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00424",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00425",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "480",
      "n": 480,
      "value_raw": "3.814130653567946",
      "value": 3.814130653567946,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00425",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00426",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "480",
      "n": 480,
      "value_raw": "3.814130653567946",
      "value": 3.814130653567946,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00426",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00427",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "4.107150890856381",
      "value": 4.107150890856381,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00427",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00428",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "4.107150890856381",
      "value": 4.107150890856381,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00428",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00429",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_response)/expected_responses",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_responses",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00429",
        "aggregate_source_id": "aggregate-source-0025",
        "source_field": "/all_records/sum(invalid_response)~1expected_responses"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00430",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "1.1262391679920256",
      "value": 1.1262391679920256,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00430",
        "aggregate_source_id": "aggregate-source-0025",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00431",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "2.693512029014528",
      "value": 2.693512029014528,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00431",
        "aggregate_source_id": "aggregate-source-0025",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00432",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "32.97560975609756",
      "value": 32.97560975609756,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00432",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/all/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00433",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "480",
      "n": 480,
      "value_raw": "35.625",
      "value": 35.625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00433",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00434",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "480",
      "n": 480,
      "value_raw": "35.625",
      "value": 35.625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00434",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00435",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "32.333333333333336",
      "value": 32.333333333333336,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00435",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00436",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "32.333333333333336",
      "value": 32.333333333333336,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00436",
        "aggregate_source_id": "aggregate-source-0024",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00437",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.49972707423580787",
      "value": 0.49972707423580787,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00437",
        "aggregate_source_id": "aggregate-source-0026",
        "source_field": "/category_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00438",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.49972707423580787",
      "value": 0.49972707423580787,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00438",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/4/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00439",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5209185376200873",
      "value": 0.5209185376200873,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00439",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/4/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00440",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_choice_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5147379912663755",
      "value": 0.5147379912663755,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00440",
        "aggregate_source_id": "aggregate-source-0026",
        "source_field": "/human_choice_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00441",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.7057000848201921",
      "value": 0.7057000848201921,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00441",
        "aggregate_source_id": "aggregate-source-0026",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00442",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00442",
        "aggregate_source_id": "aggregate-source-0026",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00443",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "1.8484486570232548",
      "value": 1.8484486570232548,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00443",
        "aggregate_source_id": "aggregate-source-0027",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00444",
      "task": "shj_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "4.6029243399971165",
      "value": 4.6029243399971165,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00444",
        "aggregate_source_id": "aggregate-source-0027",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00445",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.6498362445414847",
      "value": 0.6498362445414847,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00445",
        "aggregate_source_id": "aggregate-source-0028",
        "source_field": "/category_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00446",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.6498362445414847",
      "value": 0.6498362445414847,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00446",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/0/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00447",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.4136783296372861",
      "value": 0.4136783296372861,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00447",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/0/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00448",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_choice_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5802401746724891",
      "value": 0.5802401746724891,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00448",
        "aggregate_source_id": "aggregate-source-0028",
        "source_field": "/human_choice_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00449",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.6624112533272424",
      "value": 0.6624112533272424,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00449",
        "aggregate_source_id": "aggregate-source-0028",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00450",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00450",
        "aggregate_source_id": "aggregate-source-0028",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00451",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "1.5267412835091818",
      "value": 1.5267412835091818,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00451",
        "aggregate_source_id": "aggregate-source-0029",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00452",
      "task": "shj_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "2.809028980031144",
      "value": 2.809028980031144,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00452",
        "aggregate_source_id": "aggregate-source-0029",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00453",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5286572052401747",
      "value": 0.5286572052401747,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00453",
        "aggregate_source_id": "aggregate-source-0030",
        "source_field": "/category_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00454",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5286572052401747",
      "value": 0.5286572052401747,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00454",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/3/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00455",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.6073881519894871",
      "value": 0.6073881519894871,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00455",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/3/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00456",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_choice_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5300218340611353",
      "value": 0.5300218340611353,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00456",
        "aggregate_source_id": "aggregate-source-0030",
        "source_field": "/human_choice_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00457",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.8723015228998559",
      "value": 0.8723015228998559,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00457",
        "aggregate_source_id": "aggregate-source-0030",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00458",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00458",
        "aggregate_source_id": "aggregate-source-0030",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00459",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "1.6387252734857611",
      "value": 1.6387252734857611,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00459",
        "aggregate_source_id": "aggregate-source-0031",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00460",
      "task": "shj_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "3.189373899018392",
      "value": 3.189373899018392,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00460",
        "aggregate_source_id": "aggregate-source-0031",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00461",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5453056768558951",
      "value": 0.5453056768558951,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00461",
        "aggregate_source_id": "aggregate-source-0032",
        "source_field": "/category_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00462",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5453056768558951",
      "value": 0.5453056768558951,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00462",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/2/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00463",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.48726666143841485",
      "value": 0.48726666143841485,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00463",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/2/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00464",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_choice_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5368449781659389",
      "value": 0.5368449781659389,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00464",
        "aggregate_source_id": "aggregate-source-0032",
        "source_field": "/human_choice_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00465",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.6845411495918344",
      "value": 0.6845411495918344,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00465",
        "aggregate_source_id": "aggregate-source-0032",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00466",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00466",
        "aggregate_source_id": "aggregate-source-0032",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00467",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "1.710033063951414",
      "value": 1.710033063951414,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00467",
        "aggregate_source_id": "aggregate-source-0033",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00468",
      "task": "shj_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "3.2059191839653067",
      "value": 3.2059191839653067,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00468",
        "aggregate_source_id": "aggregate-source-0033",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00469",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.712882096069869",
      "value": 0.712882096069869,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00469",
        "aggregate_source_id": "aggregate-source-0034",
        "source_field": "/category_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00470",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.712882096069869",
      "value": 0.712882096069869,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00470",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/1/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00471",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.37513754087346596",
      "value": 0.37513754087346596,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00471",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/1/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00472",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_choice_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.5974344978165939",
      "value": 0.5974344978165939,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00472",
        "aggregate_source_id": "aggregate-source-0034",
        "source_field": "/human_choice_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00473",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.7752868790298911",
      "value": 0.7752868790298911,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00473",
        "aggregate_source_id": "aggregate-source-0034",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00474",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00474",
        "aggregate_source_id": "aggregate-source-0034",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00475",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "1.1756435394927394",
      "value": 1.1756435394927394,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00475",
        "aggregate_source_id": "aggregate-source-0035",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00476",
      "task": "shj_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch wall time. Batched throughput is not individual reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "2.271210633974988",
      "value": 2.271210633974988,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00476",
        "aggregate_source_id": "aggregate-source-0035",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00477",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6300",
      "n": 6300,
      "value_raw": "0.49603174603174605",
      "value": 0.49603174603174605,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00477",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00478",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.4988888888888889",
      "value": 0.4988888888888889,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00478",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00479",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.4938888888888889",
      "value": 0.4938888888888889,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00479",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00480",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.49603174603174605",
      "value": 0.49603174603174605,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00480",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/9/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00481",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.5733765884281666",
      "value": 0.5733765884281666,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00481",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/9/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00482",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.4879518072289157",
      "value": 0.4879518072289157,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00482",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00483",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.5107407407407407",
      "value": 0.5107407407407407,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00483",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00484",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.4930555555555556",
      "value": 0.4930555555555556,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00484",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00485",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.2013888888888889",
      "value": 0.2013888888888889,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00485",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00486",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.49805555555555553",
      "value": 0.49805555555555553,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00486",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00487",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.4583333333333333",
      "value": 0.4583333333333333,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00487",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00488",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.21875",
      "value": 0.21875,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00488",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00489",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.8499856106508037",
      "value": 0.8499856106508037,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00489",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00490",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.7777057545772214",
      "value": 0.7777057545772214,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00490",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00491",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.8541737894042561",
      "value": 0.8541737894042561,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00491",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00492",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "2.3415503592513764",
      "value": 2.3415503592513764,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00492",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00493",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.7608123578735234",
      "value": 0.7608123578735234,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00493",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00494",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.8200179098008413",
      "value": 0.8200179098008413,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00494",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00495",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "2.446572581594003",
      "value": 2.446572581594003,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00495",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00496",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00496",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00497",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00497",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00498",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00498",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00499",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00499",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00500",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00500",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00501",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00501",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00502",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00502",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00503",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.7392815189959947",
      "value": 0.7392815189959947,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00503",
        "aggregate_source_id": "aggregate-source-0037",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00504",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "1.3783955630497076",
      "value": 1.3783955630497076,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00504",
        "aggregate_source_id": "aggregate-source-0037",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00505",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "336",
      "n": 336,
      "value_raw": "2.9791666666666665",
      "value": 2.9791666666666665,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00505",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00506",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "3.2430555555555554",
      "value": 3.2430555555555554,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00506",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00507",
      "task": "medin_category_learning",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "2.78125",
      "value": 2.78125,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00507",
        "aggregate_source_id": "aggregate-source-0036",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00508",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6300",
      "n": 6300,
      "value_raw": "0.7225396825396826",
      "value": 0.7225396825396826,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00508",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00509",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.7214814814814815",
      "value": 0.7214814814814815,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00509",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00510",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.7233333333333334",
      "value": 0.7233333333333334,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00510",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00511",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.7225396825396826",
      "value": 0.7225396825396826,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00511",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/5/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00512",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.3710481759272126",
      "value": 0.3710481759272126,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00512",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/5/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00513",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.6642283419391853",
      "value": 0.6642283419391853,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00513",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00514",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.6674074074074074",
      "value": 0.6674074074074074,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00514",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00515",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.6388888888888888",
      "value": 0.6388888888888888,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00515",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00516",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.1597222222222222",
      "value": 0.1597222222222222,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00516",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00517",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.7052777777777778",
      "value": 0.7052777777777778,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00517",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00518",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.7395833333333334",
      "value": 0.7395833333333334,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00518",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00519",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.171875",
      "value": 0.171875,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00519",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00520",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.6671078363239168",
      "value": 0.6671078363239168,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00520",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00521",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.6097409588097767",
      "value": 0.6097409588097767,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00521",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00522",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.6295073289334798",
      "value": 0.6295073289334798,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00522",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00523",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "2.229662815485763",
      "value": 2.229662815485763,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00523",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00524",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.5738911586832122",
      "value": 0.5738911586832122,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00524",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00525",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.5704788409113579",
      "value": 0.5704788409113579,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00525",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00526",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "2.1745553987137267",
      "value": 2.1745553987137267,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00526",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00527",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00527",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00528",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00528",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00529",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00529",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00530",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00530",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00531",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00531",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00532",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00532",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00533",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00533",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00534",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.700797628989676",
      "value": 0.700797628989676,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00534",
        "aggregate_source_id": "aggregate-source-0039",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00535",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "1.1867500770022161",
      "value": 1.1867500770022161,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00535",
        "aggregate_source_id": "aggregate-source-0039",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00536",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "336",
      "n": 336,
      "value_raw": "2.2261904761904763",
      "value": 2.2261904761904763,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00536",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00537",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "2.3541666666666665",
      "value": 2.3541666666666665,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00537",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00538",
      "task": "medin_category_learning",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "2.1302083333333335",
      "value": 2.1302083333333335,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00538",
        "aggregate_source_id": "aggregate-source-0038",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00539",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6300",
      "n": 6300,
      "value_raw": "0.5349206349206349",
      "value": 0.5349206349206349,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00539",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00540",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.5262962962962963",
      "value": 0.5262962962962963,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00540",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00541",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.5413888888888889",
      "value": 0.5413888888888889,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00541",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00542",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.5349206349206349",
      "value": 0.5349206349206349,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00542",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/8/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00543",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.6177359556387327",
      "value": 0.6177359556387327,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00543",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/8/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00544",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.5126219162363741",
      "value": 0.5126219162363741,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00544",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00545",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.5233333333333333",
      "value": 0.5233333333333333,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00545",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00546",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.5416666666666666",
      "value": 0.5416666666666666,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00546",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00547",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.11805555555555555",
      "value": 0.11805555555555555,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00547",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00548",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.5361111111111111",
      "value": 0.5361111111111111,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00548",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00549",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.5885416666666666",
      "value": 0.5885416666666666,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00549",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00550",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.11979166666666667",
      "value": 0.11979166666666667,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00550",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00551",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.979870535451986",
      "value": 0.979870535451986,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00551",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00552",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.9344755364009023",
      "value": 0.9344755364009023,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00552",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00553",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.9862839084414294",
      "value": 0.9862839084414294,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00553",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00554",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "2.6225231696270415",
      "value": 2.6225231696270415,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00554",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00555",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.8793665606241383",
      "value": 0.8793665606241383,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00555",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00556",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.8429565283748669",
      "value": 0.8429565283748669,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00556",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00557",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "2.4028017393337366",
      "value": 2.4028017393337366,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00557",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00558",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00558",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00559",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00559",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00560",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00560",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00561",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00561",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00562",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00562",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00563",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00563",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00564",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00564",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00565",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.8493817079870496",
      "value": 0.8493817079870496,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00565",
        "aggregate_source_id": "aggregate-source-0041",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00566",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "1.383675626013428",
      "value": 1.383675626013428,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00566",
        "aggregate_source_id": "aggregate-source-0041",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00567",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "336",
      "n": 336,
      "value_raw": "2.6011904761904763",
      "value": 2.6011904761904763,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00567",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00568",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "2.6944444444444446",
      "value": 2.6944444444444446,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00568",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00569",
      "task": "medin_category_learning",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "2.53125",
      "value": 2.53125,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00569",
        "aggregate_source_id": "aggregate-source-0040",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00570",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6300",
      "n": 6300,
      "value_raw": "0.5585714285714286",
      "value": 0.5585714285714286,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00570",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00571",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.5551851851851852",
      "value": 0.5551851851851852,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00571",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00572",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.5611111111111111",
      "value": 0.5611111111111111,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00572",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00573",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.5585714285714286",
      "value": 0.5585714285714286,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00573",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/7/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00574",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.4823228014742944",
      "value": 0.4823228014742944,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00574",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/7/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00575",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.5185025817555938",
      "value": 0.5185025817555938,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00575",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00576",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.5255555555555556",
      "value": 0.5255555555555556,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00576",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00577",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.5694444444444444",
      "value": 0.5694444444444444,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00577",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00578",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.10416666666666667",
      "value": 0.10416666666666667,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00578",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00579",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.5508333333333333",
      "value": 0.5508333333333333,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00579",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00580",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.5364583333333334",
      "value": 0.5364583333333334,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00580",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00581",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.06770833333333333",
      "value": 0.06770833333333333,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00581",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00582",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.7701466953425475",
      "value": 0.7701466953425475,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00582",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00583",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.7055253255418681",
      "value": 0.7055253255418681,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00583",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00584",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.6660026077513155",
      "value": 0.6660026077513155,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00584",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00585",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "2.316152752817162",
      "value": 2.316152752817162,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00585",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00586",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.6814390154639635",
      "value": 0.6814390154639635,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00586",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00587",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.6526177012042261",
      "value": 0.6526177012042261,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00587",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00588",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "2.37828622261384",
      "value": 2.37828622261384,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00588",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00589",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00589",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00590",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00590",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00591",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00591",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00592",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00592",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00593",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00593",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00594",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00594",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00595",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00595",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00596",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.7545658399758395",
      "value": 0.7545658399758395,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00596",
        "aggregate_source_id": "aggregate-source-0043",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00597",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "1.2524018819676712",
      "value": 1.2524018819676712,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00597",
        "aggregate_source_id": "aggregate-source-0043",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00598",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "336",
      "n": 336,
      "value_raw": "3.744047619047619",
      "value": 3.744047619047619,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00598",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00599",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "3.548611111111111",
      "value": 3.548611111111111,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00599",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00600",
      "task": "medin_category_learning",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "3.890625",
      "value": 3.890625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00600",
        "aggregate_source_id": "aggregate-source-0042",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00601",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6300",
      "n": 6300,
      "value_raw": "0.5377777777777778",
      "value": 0.5377777777777778,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00601",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00602",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.5503703703703704",
      "value": 0.5503703703703704,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00602",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00603",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.5283333333333333",
      "value": 0.5283333333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00603",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00604",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.5377777777777778",
      "value": 0.5377777777777778,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00604",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/6/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00605",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6552",
      "n": 6552,
      "value_raw": "0.718356374533434",
      "value": 0.718356374533434,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00605",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/6/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00606",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.5119047619047619",
      "value": 0.5119047619047619,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00606",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00607",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0.5525925925925926",
      "value": 0.5525925925925926,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00607",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00608",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.5",
      "value": 0.5,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00608",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00609",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0.06944444444444445",
      "value": 0.06944444444444445,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00609",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00610",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0.5263888888888889",
      "value": 0.5263888888888889,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00610",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00611",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.484375",
      "value": 0.484375,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00611",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00612",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0.036458333333333336",
      "value": 0.036458333333333336,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00612",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00613",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "1.3283311508261513",
      "value": 1.3283311508261513,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00613",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00614",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "1.225972997430472",
      "value": 1.225972997430472,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00614",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00615",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "1.3538476121888101",
      "value": 1.3538476121888101,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00615",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00616",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "3.0180218445784277",
      "value": 3.0180218445784277,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00616",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00617",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "1.235200954734608",
      "value": 1.235200954734608,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00617",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00618",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "1.3984660818505765",
      "value": 1.3984660818505765,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00618",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00619",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "3.1573935623087",
      "value": 3.1573935623087,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00619",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00620",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00620",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00621",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2700",
      "n": 2700,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00621",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00622",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00622",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00623",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00623",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00624",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:learning/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3600",
      "n": 3600,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00624",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:learning~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00625",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/category_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00625",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1category_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00626",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00626",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00627",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "0.6974106660054531",
      "value": 0.6974106660054531,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00627",
        "aggregate_source_id": "aggregate-source-0045",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00628",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6972",
      "n": 6972,
      "value_raw": "1.1838492570095696",
      "value": 1.1838492570095696,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00628",
        "aggregate_source_id": "aggregate-source-0045",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00629",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "336",
      "n": 336,
      "value_raw": "5.017857142857143",
      "value": 5.017857142857143,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00629",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00630",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp1.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "144",
      "n": 144,
      "value_raw": "4.701388888888889",
      "value": 4.701388888888889,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00630",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp1.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00631",
      "task": "medin_category_learning",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "native_rating_MAE",
      "scope": "levering2020revisiting/exp2.csv:test/typicality_rating",
      "target": "original_human_numeric_response",
      "unit": "typicality_rating_steps_1_to_9",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "192",
      "n": 192,
      "value_raw": "5.255208333333333",
      "value": 5.255208333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00631",
        "aggregate_source_id": "aggregate-source-0044",
        "source_field": "/by_experiment_phase_response_kind/levering2020revisiting~1exp2.csv:test~1typicality_rating/native_rating_MAE"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00632",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.545",
      "value": 0.545,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00632",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00633",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.545",
      "value": 0.545,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00633",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/14/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00634",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.545",
      "value": 0.545,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00634",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00635",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.5437236351",
      "value": 0.5437236351,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00635",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/14/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00636",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.5375",
      "value": 0.5375,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00636",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00637",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.5375",
      "value": 0.5375,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00637",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00638",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.7185639117263178",
      "value": 0.7185639117263178,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00638",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00639",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.7185639117263178",
      "value": 0.7185639117263178,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00639",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00640",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00640",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00641",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00641",
        "aggregate_source_id": "aggregate-source-0046",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00642",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "1.36410981498193",
      "value": 1.36410981498193,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00642",
        "aggregate_source_id": "aggregate-source-0047",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00643",
      "task": "weather_prediction",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "3.116624093032442",
      "value": 3.116624093032442,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00643",
        "aggregate_source_id": "aggregate-source-0047",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00644",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.73",
      "value": 0.73,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00644",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00645",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.73",
      "value": 0.73,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00645",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/10/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00646",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.73",
      "value": 0.73,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00646",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00647",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.37561641406060986",
      "value": 0.37561641406060986,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00647",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/10/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00648",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.6275",
      "value": 0.6275,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00648",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00649",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.6275",
      "value": 0.6275,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00649",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00650",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.6508564652582025",
      "value": 0.6508564652582025,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00650",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00651",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.6508564652582025",
      "value": 0.6508564652582025,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00651",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00652",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00652",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00653",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00653",
        "aggregate_source_id": "aggregate-source-0048",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00654",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "1.1441274835087825",
      "value": 1.1441274835087825,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00654",
        "aggregate_source_id": "aggregate-source-0049",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00655",
      "task": "weather_prediction",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "2.030797475017607",
      "value": 2.030797475017607,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00655",
        "aggregate_source_id": "aggregate-source-0049",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00656",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.475",
      "value": 0.475,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00656",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00657",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.475",
      "value": 0.475,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00657",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/13/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00658",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.475",
      "value": 0.475,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00658",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00659",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.8854259830208864",
      "value": 0.8854259830208864,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00659",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/13/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00660",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.4475",
      "value": 0.4475,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00660",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00661",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.4475",
      "value": 0.4475,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00661",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00662",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "1.5703031499302507",
      "value": 1.5703031499302507,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00662",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00663",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "1.5703031499302507",
      "value": 1.5703031499302507,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00663",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00664",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00664",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00665",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00665",
        "aggregate_source_id": "aggregate-source-0050",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00666",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "1.1800554280052893",
      "value": 1.1800554280052893,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00666",
        "aggregate_source_id": "aggregate-source-0051",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00667",
      "task": "weather_prediction",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "2.0613770200288855",
      "value": 2.0613770200288855,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00667",
        "aggregate_source_id": "aggregate-source-0051",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00668",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.5375",
      "value": 0.5375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00668",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00669",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.5375",
      "value": 0.5375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00669",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/12/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00670",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.5375",
      "value": 0.5375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00670",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00671",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.49632010742722016",
      "value": 0.49632010742722016,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00671",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/12/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00672",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.56",
      "value": 0.56,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00672",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00673",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.56",
      "value": 0.56,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00673",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00674",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.692781233975655",
      "value": 0.692781233975655,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00674",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00675",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.692781233975655",
      "value": 0.692781233975655,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00675",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00676",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00676",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00677",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00677",
        "aggregate_source_id": "aggregate-source-0052",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00678",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.9030020149948541",
      "value": 0.9030020149948541,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00678",
        "aggregate_source_id": "aggregate-source-0053",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00679",
      "task": "weather_prediction",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "1.6702398169436492",
      "value": 1.6702398169436492,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00679",
        "aggregate_source_id": "aggregate-source-0053",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00680",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.505",
      "value": 0.505,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00680",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00681",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.505",
      "value": 0.505,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00681",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/11/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00682",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.505",
      "value": 0.505,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00682",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00683",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.5727370712760043",
      "value": 0.5727370712760043,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00683",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/11/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00684",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.4775",
      "value": 0.4775,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00684",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00685",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.4775",
      "value": 0.4775,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00685",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00686",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.9011097388887882",
      "value": 0.9011097388887882,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00686",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00687",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.9011097388887882",
      "value": 0.9011097388887882,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00687",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00688",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00688",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00689",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "speekenbrink2008learning/exp1.csv:learning/weather_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00689",
        "aggregate_source_id": "aggregate-source-0054",
        "source_field": "/by_experiment_phase_response_kind/speekenbrink2008learning~1exp1.csv:learning~1weather_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00690",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "1.115025101491483",
      "value": 1.115025101491483,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00690",
        "aggregate_source_id": "aggregate-source-0055",
        "source_field": "/records/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00691",
      "task": "weather_prediction",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared-device batch/scalar wall time, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "2.0054388560238294",
      "value": 2.0054388560238294,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00691",
        "aggregate_source_id": "aggregate-source-0055",
        "source_field": "/records/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00692",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.34850976167526604",
      "value": 0.34850976167526604,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00692",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00693",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.34850976167526604",
      "value": 0.34850976167526604,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00693",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00694",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.34580711005273007",
      "value": 0.34580711005273007,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00694",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00695",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.34580711005273007",
      "value": 0.34580711005273007,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00695",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00698",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00698",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00699",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00699",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00700",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.2493831610190682",
      "value": 0.2493831610190682,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00700",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00701",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.44264123000903055",
      "value": 0.44264123000903055,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00701",
        "aggregate_source_id": "aggregate-source-0056",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00702",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.7779478747330423",
      "value": 0.7779478747330423,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00702",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00703",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.7779478747330423",
      "value": 0.7779478747330423,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00703",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00704",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.733363572792047",
      "value": 0.733363572792047,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00704",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00705",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.733363572792047",
      "value": 0.733363572792047,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00705",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00708",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00708",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00709",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00709",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00710",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.23492249398259446",
      "value": 0.23492249398259446,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00710",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00711",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4101267849910073",
      "value": 0.4101267849910073,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00711",
        "aggregate_source_id": "aggregate-source-0057",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00712",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.5931847819924023",
      "value": 0.5931847819924023,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00712",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00713",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.5931847819924023",
      "value": 0.5931847819924023,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00713",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00714",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.5856627166373722",
      "value": 0.5856627166373722,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00714",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00715",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.5856627166373722",
      "value": 0.5856627166373722,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00715",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00718",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00718",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00719",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00719",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00720",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.21594438998727128",
      "value": 0.21594438998727128,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00720",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00721",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.390826279995963",
      "value": 0.390826279995963,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00721",
        "aggregate_source_id": "aggregate-source-0058",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00722",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4970853206147324",
      "value": 0.4970853206147324,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00722",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00723",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4970853206147324",
      "value": 0.4970853206147324,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00723",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00724",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4970853206147324",
      "value": 0.4970853206147324,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00724",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/19/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00725",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5556876763222045",
      "value": 0.5556876763222045,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00725",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/19/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00726",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4732379438261791",
      "value": 0.4732379438261791,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00726",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00727",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4732379438261791",
      "value": 0.4732379438261791,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00727",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00728",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.7709918455873127",
      "value": 0.7709918455873127,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00728",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00729",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.7709918455873127",
      "value": 0.7709918455873127,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00729",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00730",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00730",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00731",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00731",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00732",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.6344903849822003",
      "value": 0.6344903849822003,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00732",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00733",
      "task": "recent_probes",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "1.176500745001249",
      "value": 1.176500745001249,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00733",
        "aggregate_source_id": "aggregate-source-0059",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00734",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.8208797032326444",
      "value": 0.8208797032326444,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00734",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00735",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.8208797032326444",
      "value": 0.8208797032326444,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00735",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00736",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.8208797032326444",
      "value": 0.8208797032326444,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00736",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/15/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00737",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.23578292864535969",
      "value": 0.23578292864535969,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00737",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/15/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00738",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.7795442501324854",
      "value": 0.7795442501324854,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00738",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00739",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.7795442501324854",
      "value": 0.7795442501324854,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00739",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00740",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4578059995059755",
      "value": 0.4578059995059755,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00740",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00741",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4578059995059755",
      "value": 0.4578059995059755,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00741",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00742",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00742",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00743",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00743",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00744",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.6256160135089885",
      "value": 0.6256160135089885,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00744",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00745",
      "task": "recent_probes",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "1.0149569410132244",
      "value": 1.0149569410132244,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00745",
        "aggregate_source_id": "aggregate-source-0060",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00746",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.555908850026497",
      "value": 0.555908850026497,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00746",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00747",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.555908850026497",
      "value": 0.555908850026497,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00747",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00748",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.555908850026497",
      "value": 0.555908850026497,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00748",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/18/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00749",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.6298152518148116",
      "value": 0.6298152518148116,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00749",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/18/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00750",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5310015898251192",
      "value": 0.5310015898251192,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00750",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00751",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5310015898251192",
      "value": 0.5310015898251192,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00751",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00752",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "1.005636722654248",
      "value": 1.005636722654248,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00752",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00753",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "1.005636722654248",
      "value": 1.005636722654248,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00753",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00754",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00754",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00755",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00755",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00756",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5566262754728086",
      "value": 0.5566262754728086,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00756",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00757",
      "task": "recent_probes",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.9823811170062982",
      "value": 0.9823811170062982,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00757",
        "aggregate_source_id": "aggregate-source-0061",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00758",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.49470058293587704",
      "value": 0.49470058293587704,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00758",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00759",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.49470058293587704",
      "value": 0.49470058293587704,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00759",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00760",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.49470058293587704",
      "value": 0.49470058293587704,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00760",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/17/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00761",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.6114059161122953",
      "value": 0.6114059161122953,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00761",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/17/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00762",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4692633810280869",
      "value": 0.4692633810280869,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00762",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00763",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.4692633810280869",
      "value": 0.4692633810280869,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00763",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00764",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.8513555441647751",
      "value": 0.8513555441647751,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00764",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00765",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.8513555441647751",
      "value": 0.8513555441647751,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00765",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00766",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00766",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00767",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00767",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00768",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5662226635031402",
      "value": 0.5662226635031402,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00768",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00769",
      "task": "recent_probes",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.9796566769946367",
      "value": 0.9796566769946367,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00769",
        "aggregate_source_id": "aggregate-source-0062",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00770",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5349761526232114",
      "value": 0.5349761526232114,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00770",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00771",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5349761526232114",
      "value": 0.5349761526232114,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00771",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00772",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5349761526232114",
      "value": 0.5349761526232114,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00772",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/16/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00773",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.7306563460821602",
      "value": 0.7306563460821602,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00773",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/16/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00774",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5551139374668786",
      "value": 0.5551139374668786,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00774",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00775",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5551139374668786",
      "value": 0.5551139374668786,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00775",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/human_response_agreement"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00776",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "1.2047282841900864",
      "value": 1.2047282841900864,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00776",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00777",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "1.2047282841900864",
      "value": 1.2047282841900864,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00777",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/typed_choice_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00778",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00778",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00779",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "enkavi2019recentprobes/exp1.csv:original_session/recent_probes_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00779",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/by_experiment_phase_response_kind/enkavi2019recentprobes~1exp1.csv:original_session~1recent_probes_choice/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00780",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.5928663649829105",
      "value": 0.5928663649829105,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00780",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00781",
      "task": "recent_probes",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.9662066000164486",
      "value": 0.9662066000164486,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00781",
        "aggregate_source_id": "aggregate-source-0063",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v00782",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.25468547878350717",
      "value": 0.25468547878350717,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00782",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00783",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.25474360984789085",
      "value": 0.25474360984789085,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00783",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00784",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.2553055385335052",
      "value": 0.2553055385335052,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00784",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00785",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2553708640426533",
      "value": 0.2553708640426533,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00785",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00786",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum(p-onehot_ground_key)^2)",
      "scope": "generative",
      "target": "original_hidden_rule_ground_key",
      "unit": "sum_over_four_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.7996084660177191",
      "value": 0.7996084660177191,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00786",
        "aggregate_source_id": "aggregate-source-0065",
        "source_field": "/objective_labelled_trials/mean(sum(p-onehot_ground_key)^2)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00787",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum(p-onehot_ground_key)^2)",
      "scope": "predictive",
      "target": "original_hidden_rule_ground_key",
      "unit": "sum_over_four_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.7855621017209368",
      "value": 0.7855621017209368,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00787",
        "aggregate_source_id": "aggregate-source-0066",
        "source_field": "/objective_labelled_trials/mean(sum(p-onehot_ground_key)^2)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00788",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.4615976053331359",
      "value": 1.4615976053331359,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00788",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00789",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.4617539575177372",
      "value": 1.4617539575177372,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00789",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00790",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_outputs)/sum(trials)",
      "scope": "generative",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_trial_denominator",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00790",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/per_participant/generative/sum(invalid_outputs)~1sum(trials)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00791",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_outputs)/sum(trials)",
      "scope": "predictive",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_trial_denominator",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00791",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/per_participant/predictive/sum(invalid_outputs)~1sum(trials)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00792",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "generative",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.22769570400123484",
      "value": 0.22769570400123484,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00792",
        "aggregate_source_id": "aggregate-source-0065",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00793",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "predictive",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.22879733549780212",
      "value": 0.22879733549780212,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00793",
        "aggregate_source_id": "aggregate-source-0066",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00794",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "generative",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.49231341798440553",
      "value": 0.49231341798440553,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00794",
        "aggregate_source_id": "aggregate-source-0065",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00795",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "predictive",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.4946400350017939",
      "value": 0.4946400350017939,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00795",
        "aggregate_source_id": "aggregate-source-0066",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00796",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.15242601496484542",
      "value": 0.15242601496484542,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00796",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00797",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.15234436255292458",
      "value": 0.15234436255292458,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00797",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00798",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.15157588425691806",
      "value": 0.15157588425691806,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00798",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00799",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.15148188803512624",
      "value": 0.15148188803512624,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00799",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00800",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.1816453139641607",
      "value": 0.1816453139641607,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00800",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00801",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.18182530970675867",
      "value": 0.18182530970675867,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00801",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00802",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.1814647419677628",
      "value": 0.1814647419677628,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00802",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00803",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.18166849615806804",
      "value": 0.18166849615806804,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00803",
        "aggregate_source_id": "aggregate-source-0064",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00804",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.3139030190672591",
      "value": 0.3139030190672591,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00804",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00805",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.3143327583503215",
      "value": 0.3143327583503215,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00805",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00806",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.3255558582025881",
      "value": 0.3255558582025881,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00806",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00807",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.3260153677277717",
      "value": 0.3260153677277717,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00807",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00808",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum(p-onehot_ground_key)^2)",
      "scope": "generative",
      "target": "original_hidden_rule_ground_key",
      "unit": "sum_over_four_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.1555367489236827",
      "value": 1.1555367489236827,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00808",
        "aggregate_source_id": "aggregate-source-0068",
        "source_field": "/objective_labelled_trials/mean(sum(p-onehot_ground_key)^2)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00809",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum(p-onehot_ground_key)^2)",
      "scope": "predictive",
      "target": "original_hidden_rule_ground_key",
      "unit": "sum_over_four_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.8041798005708732",
      "value": 0.8041798005708732,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00809",
        "aggregate_source_id": "aggregate-source-0069",
        "source_field": "/objective_labelled_trials/mean(sum(p-onehot_ground_key)^2)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00810",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.4307050334551046",
      "value": 1.4307050334551046,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00810",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00811",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.4306021561560678",
      "value": 1.4306021561560678,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00811",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00812",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll_original_full_vocab",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "human_nll_original_full_vocab",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.4436930767424163",
      "value": 1.4436930767424163,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00812",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/human_nll_original_full_vocab/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00813",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll_original_full_vocab",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "human_nll_original_full_vocab",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.4434991519204758",
      "value": 1.4434991519204758,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00813",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/human_nll_original_full_vocab/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00814",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_outputs)/sum(trials)",
      "scope": "generative",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_trial_denominator",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.003998745491610475",
      "value": 0.003998745491610475,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00814",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/per_participant/generative/sum(invalid_outputs)~1sum(trials)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00815",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_outputs)/sum(trials)",
      "scope": "predictive",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_trial_denominator",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.005880508075897757",
      "value": 0.005880508075897757,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00815",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/per_participant/predictive/sum(invalid_outputs)~1sum(trials)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00816",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "generative",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.49588187399785966",
      "value": 0.49588187399785966,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00816",
        "aggregate_source_id": "aggregate-source-0068",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00817",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "predictive",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.8674724925076589",
      "value": 0.8674724925076589,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00817",
        "aggregate_source_id": "aggregate-source-0069",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00818",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "generative",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.3402204730082303",
      "value": 1.3402204730082303,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00818",
        "aggregate_source_id": "aggregate-source-0068",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00819",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "predictive",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.7997518350020982",
      "value": 1.7997518350020982,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00819",
        "aggregate_source_id": "aggregate-source-0069",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00820",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.288157966810707",
      "value": 0.288157966810707,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00820",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00821",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2902618786263133",
      "value": 0.2902618786263133,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00821",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00822",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.22675895732209633",
      "value": 0.22675895732209633,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00822",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00823",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.22667398463227223",
      "value": 0.22667398463227223,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00823",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00824",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.13779026430255414",
      "value": 0.13779026430255414,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00824",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00825",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.13721185510428102",
      "value": 0.13721185510428102,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00825",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00826",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.17231576372569007",
      "value": 0.17231576372569007,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00826",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00827",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.17304375098008468",
      "value": 0.17304375098008468,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00827",
        "aggregate_source_id": "aggregate-source-0067",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00828",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.28155142111511694",
      "value": 0.28155142111511694,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00828",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00829",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2816371334483299",
      "value": 0.2816371334483299,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00829",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00830",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.3006092268765138",
      "value": 0.3006092268765138,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00830",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00831",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.300768386388584",
      "value": 0.300768386388584,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00831",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00832",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum(p-onehot_ground_key)^2)",
      "scope": "generative",
      "target": "original_hidden_rule_ground_key",
      "unit": "sum_over_four_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.9132272108344495",
      "value": 0.9132272108344495,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00832",
        "aggregate_source_id": "aggregate-source-0071",
        "source_field": "/objective_labelled_trials/mean(sum(p-onehot_ground_key)^2)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00833",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum(p-onehot_ground_key)^2)",
      "scope": "predictive",
      "target": "original_hidden_rule_ground_key",
      "unit": "sum_over_four_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.7581845016046018",
      "value": 0.7581845016046018,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00833",
        "aggregate_source_id": "aggregate-source-0072",
        "source_field": "/objective_labelled_trials/mean(sum(p-onehot_ground_key)^2)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00834",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.3435348184735838",
      "value": 1.3435348184735838,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00834",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00835",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.3423392670155254",
      "value": 1.3423392670155254,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00835",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00836",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll_original_full_vocab",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "human_nll_original_full_vocab",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.3436881797270472",
      "value": 1.3436881797270472,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00836",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/human_nll_original_full_vocab/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00837",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll_original_full_vocab",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "human_nll_original_full_vocab",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.3424915100164654",
      "value": 1.3424915100164654,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00837",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/human_nll_original_full_vocab/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00838",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_outputs)/sum(trials)",
      "scope": "generative",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_trial_denominator",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00838",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/per_participant/generative/sum(invalid_outputs)~1sum(trials)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00839",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_outputs)/sum(trials)",
      "scope": "predictive",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_trial_denominator",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00839",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/per_participant/predictive/sum(invalid_outputs)~1sum(trials)"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00840",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "generative",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.5838313569984166",
      "value": 0.5838313569984166,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00840",
        "aggregate_source_id": "aggregate-source-0071",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00841",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "predictive",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.6991580895119114",
      "value": 0.6991580895119114,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00841",
        "aggregate_source_id": "aggregate-source-0072",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00842",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "generative",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.0767154839995783",
      "value": 1.0767154839995783,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00842",
        "aggregate_source_id": "aggregate-source-0071",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00843",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "predictive",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "All 75 participants and 12754 trial requests, includes first call; source adapter timing.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.4092514920048416",
      "value": 1.4092514920048416,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00843",
        "aggregate_source_id": "aggregate-source-0072",
        "source_field": "/all_trials/seconds_shared_gpu/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00844",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.21845648433867665",
      "value": 0.21845648433867665,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00844",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00845",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.21804923945428886",
      "value": 0.21804923945428886,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00845",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00846",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.21554997964385",
      "value": 0.21554997964385,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00846",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00847",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.21554022267523915",
      "value": 0.21554022267523915,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00847",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00848",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.14605572737165445",
      "value": 0.14605572737165445,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00848",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00849",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.14685588834875332",
      "value": 0.14685588834875332,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00849",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00850",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.16671230140964605",
      "value": 0.16671230140964605,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00850",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00851",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.1670064293554963",
      "value": 0.1670064293554963,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00851",
        "aggregate_source_id": "aggregate-source-0070",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00852",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.5537500000000001",
      "value": 0.5537500000000001,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00852",
        "aggregate_source_id": "aggregate-source-0073",
        "source_field": "/accuracy_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00853",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_pre_step_additional",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.54875",
      "value": 0.54875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00853",
        "aggregate_source_id": "aggregate-source-0073",
        "source_field": "/accuracy_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00854",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_post_step_original",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.291870375",
      "value": 0.291870375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00854",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/0/event_brier_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00855",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_pre_step_additional",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.287845375",
      "value": 0.287845375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00855",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/0/event_brier_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00856",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_original",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_post_step_original",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.291870375",
      "value": 0.291870375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00856",
        "aggregate_source_id": "aggregate-source-0073",
        "source_field": "/qsr_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00857",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_pre_step_additional",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_pre_step_additional",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.28784537499999996",
      "value": 0.28784537499999996,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00857",
        "aggregate_source_id": "aggregate-source-0073",
        "source_field": "/qsr_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00858",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.805159215349704",
      "value": 0.805159215349704,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00858",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/0/latency_p50_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00859",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "1.4069211918860671",
      "value": 1.4069211918860671,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00859",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/0/latency_p95_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00860",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_mean_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "9.025",
      "value": 9.025,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00860",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/0/pseudo_regret_pre_step_mean_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00861",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_total_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "7220.0",
      "value": 7220.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00861",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/0/pseudo_regret_pre_step_total_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00862",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_original",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "50.992875",
      "value": 50.992875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00862",
        "aggregate_source_id": "aggregate-source-0073",
        "source_field": "/mean_reward_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00863",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.6562499999999999",
      "value": 0.6562499999999999,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00863",
        "aggregate_source_id": "aggregate-source-0075",
        "source_field": "/accuracy_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00864",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_pre_step_additional",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.69125",
      "value": 0.69125,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00864",
        "aggregate_source_id": "aggregate-source-0075",
        "source_field": "/accuracy_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00865",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_post_step_original",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.24787287500000002",
      "value": 0.24787287500000002,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00865",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/3/event_brier_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00866",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_pre_step_additional",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.233697875",
      "value": 0.233697875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00866",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/3/event_brier_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00867",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_original",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_post_step_original",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.247872875",
      "value": 0.247872875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00867",
        "aggregate_source_id": "aggregate-source-0075",
        "source_field": "/qsr_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00868",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_pre_step_additional",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_pre_step_additional",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.23369787500000005",
      "value": 0.23369787500000005,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00868",
        "aggregate_source_id": "aggregate-source-0075",
        "source_field": "/qsr_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00869",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.9019797759538051",
      "value": 0.9019797759538051,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00869",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/3/latency_p50_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00870",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "1.2807384535350137",
      "value": 1.2807384535350137,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00870",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/3/latency_p95_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00871",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_mean_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "6.175",
      "value": 6.175,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00871",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/3/pseudo_regret_pre_step_mean_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00872",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_total_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "4940.0",
      "value": 4940.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00872",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/3/pseudo_regret_pre_step_total_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00873",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_original",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "53.84287500000001",
      "value": 53.84287500000001,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00873",
        "aggregate_source_id": "aggregate-source-0075",
        "source_field": "/mean_reward_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00874",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.57125",
      "value": 0.57125,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00874",
        "aggregate_source_id": "aggregate-source-0076",
        "source_field": "/accuracy_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00875",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_pre_step_additional",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.58375",
      "value": 0.58375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00875",
        "aggregate_source_id": "aggregate-source-0076",
        "source_field": "/accuracy_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00876",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_post_step_original",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.2998905",
      "value": 0.2998905,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00876",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/1/event_brier_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00877",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_pre_step_additional",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.29461550000000003",
      "value": 0.29461550000000003,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00877",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/1/event_brier_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00878",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_original",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_post_step_original",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.29989050000000006",
      "value": 0.29989050000000006,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00878",
        "aggregate_source_id": "aggregate-source-0076",
        "source_field": "/qsr_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00879",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_pre_step_additional",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_pre_step_additional",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.29461550000000003",
      "value": 0.29461550000000003,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00879",
        "aggregate_source_id": "aggregate-source-0076",
        "source_field": "/qsr_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00880",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.25082153900075355",
      "value": 0.25082153900075355,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00880",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/1/latency_p50_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00881",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.3917227297515637",
      "value": 0.3917227297515637,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00881",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/1/latency_p95_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00882",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_mean_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "8.325",
      "value": 8.325,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00882",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/1/pseudo_regret_pre_step_mean_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00883",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_total_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "6660.0",
      "value": 6660.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00883",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/1/pseudo_regret_pre_step_total_additional"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00884",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_original",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "51.692875",
      "value": 51.692875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00884",
        "aggregate_source_id": "aggregate-source-0076",
        "source_field": "/mean_reward_original"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v00885",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.48999999999999994",
      "value": 0.48999999999999994,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00885",
        "aggregate_source_id": "aggregate-source-0077",
        "source_field": "/accuracy_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00886",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_pre_step_additional",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.49",
      "value": 0.49,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00886",
        "aggregate_source_id": "aggregate-source-0077",
        "source_field": "/accuracy_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00887",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_post_step_original",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.25",
      "value": 0.25,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00887",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/2/event_brier_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00888",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_pre_step_additional",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.25",
      "value": 0.25,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00888",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/2/event_brier_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00889",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_original",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_post_step_original",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.25",
      "value": 0.25,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00889",
        "aggregate_source_id": "aggregate-source-0077",
        "source_field": "/qsr_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00890",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_pre_step_additional",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_pre_step_additional",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.25",
      "value": 0.25,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00890",
        "aggregate_source_id": "aggregate-source-0077",
        "source_field": "/qsr_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00891",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.28379604796646163",
      "value": 0.28379604796646163,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00891",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/2/latency_p50_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00892",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "1.1930034184915699",
      "value": 1.1930034184915699,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00892",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/2/latency_p95_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00893",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_mean_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "10.2",
      "value": 10.2,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00893",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/2/pseudo_regret_pre_step_mean_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00894",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_total_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "8160.0",
      "value": 8160.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00894",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/2/pseudo_regret_pre_step_total_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00895",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_original",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "49.817875",
      "value": 49.817875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00895",
        "aggregate_source_id": "aggregate-source-0077",
        "source_field": "/mean_reward_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00896",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.505",
      "value": 0.505,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00896",
        "aggregate_source_id": "aggregate-source-0078",
        "source_field": "/accuracy_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00897",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_pre_step_additional",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.52",
      "value": 0.52,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00897",
        "aggregate_source_id": "aggregate-source-0078",
        "source_field": "/accuracy_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00898",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_post_step_original",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.423496875",
      "value": 0.423496875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00898",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/4/event_brier_post_step_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00899",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "event_brier_pre_step_additional",
      "scope": "all",
      "target": "Independently reported confidence that the chosen machine is currently best, paired with latent-mean best-arm label; never action policy probability.",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.407746875",
      "value": 0.407746875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00899",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/4/event_brier_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00900",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_original",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_post_step_original",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.423496875",
      "value": 0.423496875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00900",
        "aggregate_source_id": "aggregate-source-0078",
        "source_field": "/qsr_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00901",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "event_answer_brier",
      "metric_variant": "1_minus_qsr_pre_step_additional",
      "scope": "all",
      "target": "independent_confidence_selected_arm_is_best_pre_step_additional",
      "unit": "binary_single_event_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Derived exactly as 1 − QSR; source zero-confidence wording is guessing, not a universal calibration claim.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.40774687499999995",
      "value": 0.40774687499999995,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00901",
        "aggregate_source_id": "aggregate-source-0078",
        "source_field": "/qsr_pre_step_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00902",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.9349961949919816",
      "value": 0.9349961949919816,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00902",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/4/latency_p50_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00903",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_two_queries_seconds",
      "scope": "all",
      "target": "two_requests_per_environment_trial",
      "unit": "seconds_per_action_plus_independent_confidence",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared device; no exclusive-device performance claim.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "1.3409819887630874",
      "value": 1.3409819887630874,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00903",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/4/latency_p95_two_queries_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00904",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_mean_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "9.6",
      "value": 9.6,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00904",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/4/pseudo_regret_pre_step_mean_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00905",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_pre_step_total_additional",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "7680.0",
      "value": 7680.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00905",
        "aggregate_source_id": "aggregate-source-0074",
        "source_field": "/models/4/pseudo_regret_pre_step_total_additional"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00906",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_original",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "800",
      "n": 800,
      "value_raw": "50.417874999999995",
      "value": 50.417874999999995,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00906",
        "aggregate_source_id": "aggregate-source-0078",
        "source_field": "/mean_reward_original"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00907",
      "task": "instrumental_learning",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00907",
        "aggregate_source_id": "aggregate-source-0079",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00908",
      "task": "instrumental_learning",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.21156837948365137",
      "value": 0.21156837948365137,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00908",
        "aggregate_source_id": "aggregate-source-0079",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00909",
      "task": "instrumental_learning",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.3680958074692171",
      "value": 0.3680958074692171,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00909",
        "aggregate_source_id": "aggregate-source-0079",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00910",
      "task": "instrumental_learning",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_total",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "125.5",
      "value": 125.5,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00910",
        "aggregate_source_id": "aggregate-source-0079",
        "source_field": "/pseudo_regret_total"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00911",
      "task": "instrumental_learning",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.48333333333333334",
      "value": 0.48333333333333334,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00911",
        "aggregate_source_id": "aggregate-source-0079",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00912",
      "task": "instrumental_learning",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00912",
        "aggregate_source_id": "aggregate-source-0080",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00913",
      "task": "instrumental_learning",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.8358070860267617",
      "value": 0.8358070860267617,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00913",
        "aggregate_source_id": "aggregate-source-0080",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00914",
      "task": "instrumental_learning",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "1.2889074347942369",
      "value": 1.2889074347942369,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00914",
        "aggregate_source_id": "aggregate-source-0080",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00915",
      "task": "instrumental_learning",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_total",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "89.5",
      "value": 89.5,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00915",
        "aggregate_source_id": "aggregate-source-0080",
        "source_field": "/pseudo_regret_total"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00916",
      "task": "instrumental_learning",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.528125",
      "value": 0.528125,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00916",
        "aggregate_source_id": "aggregate-source-0080",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00917",
      "task": "instrumental_learning",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00917",
        "aggregate_source_id": "aggregate-source-0081",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00918",
      "task": "instrumental_learning",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.33241020250716247",
      "value": 0.33241020250716247,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00918",
        "aggregate_source_id": "aggregate-source-0081",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00919",
      "task": "instrumental_learning",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.520885564587661",
      "value": 0.520885564587661,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00919",
        "aggregate_source_id": "aggregate-source-0081",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00920",
      "task": "instrumental_learning",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_total",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "100.0",
      "value": 100.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00920",
        "aggregate_source_id": "aggregate-source-0081",
        "source_field": "/pseudo_regret_total"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00921",
      "task": "instrumental_learning",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.49166666666666664",
      "value": 0.49166666666666664,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00921",
        "aggregate_source_id": "aggregate-source-0081",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00922",
      "task": "instrumental_learning",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00922",
        "aggregate_source_id": "aggregate-source-0082",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00923",
      "task": "instrumental_learning",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.39689242499298416",
      "value": 0.39689242499298416,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00923",
        "aggregate_source_id": "aggregate-source-0082",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00924",
      "task": "instrumental_learning",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.6334703164873644",
      "value": 0.6334703164873644,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00924",
        "aggregate_source_id": "aggregate-source-0082",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00925",
      "task": "instrumental_learning",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_total",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "64.0",
      "value": 64.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00925",
        "aggregate_source_id": "aggregate-source-0082",
        "source_field": "/pseudo_regret_total"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00926",
      "task": "instrumental_learning",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.5354166666666667",
      "value": 0.5354166666666667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00926",
        "aggregate_source_id": "aggregate-source-0082",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00927",
      "task": "instrumental_learning",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00927",
        "aggregate_source_id": "aggregate-source-0083",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00928",
      "task": "instrumental_learning",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.4501309310144279",
      "value": 0.4501309310144279,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00928",
        "aggregate_source_id": "aggregate-source-0083",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00929",
      "task": "instrumental_learning",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.6639741703373147",
      "value": 0.6639741703373147,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00929",
        "aggregate_source_id": "aggregate-source-0083",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00930",
      "task": "instrumental_learning",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_total",
      "scope": "all",
      "target": "latent_expected_reward_gap",
      "unit": "task_specific_reward_units",
      "aggregation": "sum_over_trials",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "96.0",
      "value": 96.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00930",
        "aggregate_source_id": "aggregate-source-0083",
        "source_field": "/pseudo_regret_total"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00931",
      "task": "instrumental_learning",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.5302083333333333",
      "value": 0.5302083333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00931",
        "aggregate_source_id": "aggregate-source-0083",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00932",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "mean_optimal_choice_rate",
      "scope": "all",
      "target": "latent_mean_optimal_arm",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.5479838709677419",
      "value": 0.5479838709677419,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00932",
        "aggregate_source_id": "aggregate-source-0084",
        "source_field": "/mean_optimal_choice_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00933",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.7022374780763286",
      "value": 0.7022374780763286,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00933",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/0/legal_nll_eps1e8"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00934",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_conditioned_nll_eps1e8",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Explicit epsilon 1e-8 legal-response NLL.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.7022374780763286",
      "value": 0.7022374780763286,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00934",
        "aggregate_source_id": "aggregate-source-0086",
        "source_field": "/legal_conditioned_nll_eps1e8"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00935",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "native_invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00935",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/0/native_invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00936",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "shared with this project Centaur70B evaluation; not exclusive GPU latency; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.0405014064162969",
      "value": 0.0405014064162969,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00936",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/0/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00937",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "shared with this project Centaur70B evaluation; not exclusive GPU latency; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.051635894505307055",
      "value": 0.051635894505307055,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00937",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/0/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00938",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_free_additional_metric",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "50.97992511520737",
      "value": 50.97992511520737,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00938",
        "aggregate_source_id": "aggregate-source-0084",
        "source_field": "/mean_reward_free_additional_metric"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00939",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.5178796649816175",
      "value": 0.5178796649816175,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00939",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/1/legal_nll_eps1e8"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00940",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Epsilon and response normalization follow source summary.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.5178796649816175",
      "value": 0.5178796649816175,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00940",
        "aggregate_source_id": "aggregate-source-0087",
        "source_field": "/mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00941",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "native_invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00941",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/1/native_invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00942",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.23156135197496042",
      "value": 0.23156135197496042,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00942",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00943",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.2565947856201092",
      "value": 0.2565947856201092,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00943",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00944",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.5031625522657853",
      "value": 0.5031625522657853,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00944",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/2/legal_nll_eps1e8"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00945",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "original_full_vocabulary_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.5031676156180246",
      "value": 0.5031676156180246,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00945",
        "aggregate_source_id": "aggregate-source-0088",
        "source_field": "/published_full_vocabulary_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00946",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "native_invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00946",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/2/native_invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00947",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_free_additional_metric",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "52.89032258064516",
      "value": 52.89032258064516,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00947",
        "aggregate_source_id": "aggregate-source-0089",
        "source_field": "/mean_reward_free_additional_metric"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00948",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.4093538186756058",
      "value": 0.4093538186756058,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00948",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/3/legal_nll_eps1e8"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00949",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "original_full_vocabulary_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.40978121076311386",
      "value": 0.40978121076311386,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00949",
        "aggregate_source_id": "aggregate-source-0090",
        "source_field": "/published_full_vocabulary_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00950",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "native_invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00950",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/3/native_invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00951",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward_free_additional_metric",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "53.89014976958526",
      "value": 53.89014976958526,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00951",
        "aggregate_source_id": "aggregate-source-0091",
        "source_field": "/mean_reward_free_additional_metric"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00952",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.6613339370488202",
      "value": 0.6613339370488202,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00952",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/4/legal_nll_eps1e8"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00953",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Epsilon and response normalization follow source summary.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.6613339370488203",
      "value": 0.6613339370488203,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00953",
        "aggregate_source_id": "aggregate-source-0092",
        "source_field": "/mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00954",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "native_invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00954",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/4/native_invalid_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00955",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.15493965800851583",
      "value": 0.15493965800851583,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00955",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/4/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00956",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.1770249250170309",
      "value": 0.1770249250170309,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00956",
        "aggregate_source_id": "aggregate-source-0085",
        "source_field": "/models/4/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v00957",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "invalid_rate",
      "metric_variant": "count(chosen_action_not_legal)/all_steps",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_steps",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00957",
        "aggregate_source_id": "aggregate-source-0093",
        "source_field": "/all_steps/count(chosen_action_not_legal)~1all_steps"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00958",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Engineering request distribution over all16 conditions, not a condition-matched speed ranking.; Shared GPU/CPU; elapsed time is not exclusive throughput.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.09746805350005161",
      "value": 0.09746805350005161,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00958",
        "aggregate_source_id": "aggregate-source-0093",
        "source_field": "/all_steps/inference_seconds/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00959",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Engineering request distribution over all16 conditions, not a condition-matched speed ranking.; Shared GPU/CPU; elapsed time is not exclusive throughput.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.27857955299987225",
      "value": 0.27857955299987225,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00959",
        "aggregate_source_id": "aggregate-source-0093",
        "source_field": "/all_steps/inference_seconds/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00960",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "9.35",
      "value": 9.35,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00960",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/0/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00961",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00961",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/1/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00962",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00962",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/2/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00963",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "6638.312778338334",
      "value": 6638.312778338334,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00963",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/3/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00964",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00964",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/4/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00965",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "748.0705240486853",
      "value": 748.0705240486853,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00965",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/5/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00966",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00966",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/6/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00967",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "605.1677074621782",
      "value": 605.1677074621782,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00967",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/7/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00968",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00968",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/8/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00969",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00969",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/9/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00970",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00970",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/10/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00971",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "13056.234530842492",
      "value": 13056.234530842492,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00971",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/11/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00972",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00972",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/12/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00973",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "669.2421252935429",
      "value": 669.2421252935429,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00973",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/13/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00974",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00974",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/14/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00975",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "9627.98677206628",
      "value": 9627.98677206628,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00975",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/15/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00976",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "214.9",
      "value": 214.9,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00976",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/0/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00977",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "268.27395003692266",
      "value": 268.27395003692266,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00977",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/1/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00978",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "399.1",
      "value": 399.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00978",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/2/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00979",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-1245.1652191889334",
      "value": -1245.1652191889334,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00979",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/3/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00980",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "75.9",
      "value": 75.9,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00980",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/4/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00981",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-140.0488254976331",
      "value": -140.0488254976331,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00981",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/5/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00982",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "401.1",
      "value": 401.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00982",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/6/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00983",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4392.256327747347",
      "value": 4392.256327747347,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00983",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/7/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00984",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "72.5",
      "value": 72.5,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00984",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/8/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00985",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "487.39991791338514",
      "value": 487.39991791338514,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00985",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/9/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00986",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "402.8",
      "value": 402.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00986",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/10/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00987",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-7654.0736853464905",
      "value": -7654.0736853464905,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00987",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/11/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00988",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "73.8",
      "value": 73.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00988",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/12/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00989",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-269.59091798551236",
      "value": -269.59091798551236,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00989",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/13/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00990",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "400.8",
      "value": 400.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00990",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/14/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00991",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-5097.250683553856",
      "value": -5097.250683553856,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00991",
        "aggregate_source_id": "aggregate-source-0094",
        "source_field": "/conditions/15/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00992",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "invalid_rate",
      "metric_variant": "count(chosen_action_not_legal)/all_steps",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_steps",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00992",
        "aggregate_source_id": "aggregate-source-0095",
        "source_field": "/all_steps/count(chosen_action_not_legal)~1all_steps"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00993",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Engineering request distribution over all16 conditions, not a condition-matched speed ranking.; Shared GPU/CPU; elapsed time is not exclusive throughput.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.13460791400029848",
      "value": 0.13460791400029848,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00993",
        "aggregate_source_id": "aggregate-source-0095",
        "source_field": "/all_steps/inference_seconds/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00994",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Engineering request distribution over all16 conditions, not a condition-matched speed ranking.; Shared GPU/CPU; elapsed time is not exclusive throughput.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.2532462140006828",
      "value": 0.2532462140006828,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00994",
        "aggregate_source_id": "aggregate-source-0095",
        "source_field": "/all_steps/inference_seconds/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00995",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00995",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/0/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00996",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "662.5539294832477",
      "value": 662.5539294832477,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00996",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/1/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00997",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00997",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/2/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00998",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "6967.367472604819",
      "value": 6967.367472604819,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00998",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/3/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v00999",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v00999",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/4/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01000",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "411.19517744532027",
      "value": 411.19517744532027,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01000",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/5/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01001",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "80.38666666666553",
      "value": 80.38666666666553,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01001",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/6/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01002",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "878.253746178623",
      "value": 878.253746178623,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01002",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/7/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01003",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01003",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/8/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01004",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "429.57018073498847",
      "value": 429.57018073498847,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01004",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/9/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01005",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01005",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/10/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01006",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "8617.935119897127",
      "value": 8617.935119897127,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01006",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/11/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01007",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01007",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/12/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01008",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "419.7597688033603",
      "value": 419.7597688033603,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01008",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/13/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01009",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01009",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/14/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01010",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "2079.5728078748175",
      "value": 2079.5728078748175,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01010",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/15/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01011",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "76.0",
      "value": 76.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01011",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/0/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01012",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-394.2799794463209",
      "value": -394.2799794463209,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01012",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/1/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01013",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "399.1",
      "value": 399.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01013",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/2/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01014",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-1574.219913455496",
      "value": -1574.219913455496,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01014",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/3/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01015",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "75.9",
      "value": 75.9,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01015",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/4/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01016",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "196.82652110573173",
      "value": 196.82652110573173,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01016",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/5/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01017",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "522.4666666666667",
      "value": 522.4666666666667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01017",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/6/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01018",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4119.170289030897",
      "value": 4119.170289030897,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01018",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/7/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01019",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "72.5",
      "value": 72.5,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01019",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/8/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01020",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "57.82973717839446",
      "value": 57.82973717839446,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01020",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/9/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01021",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "402.8",
      "value": 402.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01021",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/10/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01022",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-3215.7742744015563",
      "value": -3215.7742744015563,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01022",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/11/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01023",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "73.8",
      "value": 73.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01023",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/12/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01024",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-20.10856149532947",
      "value": -20.10856149532947,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01024",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/13/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01025",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "400.8",
      "value": 400.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01025",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/14/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01026",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "2451.1632806376174",
      "value": 2451.1632806376174,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01026",
        "aggregate_source_id": "aggregate-source-0096",
        "source_field": "/conditions/15/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01027",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "invalid_rate",
      "metric_variant": "count(chosen_action_not_legal)/all_steps",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_steps",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01027",
        "aggregate_source_id": "aggregate-source-0097",
        "source_field": "/all_steps/count(chosen_action_not_legal)~1all_steps"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01028",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Engineering request distribution over all16 conditions, not a condition-matched speed ranking.; Shared GPU/CPU; elapsed time is not exclusive throughput.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.13141860549512785",
      "value": 0.13141860549512785,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01028",
        "aggregate_source_id": "aggregate-source-0097",
        "source_field": "/all_steps/inference_seconds/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01029",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Engineering request distribution over all16 conditions, not a condition-matched speed ranking.; Shared GPU/CPU; elapsed time is not exclusive throughput.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.22186691498791333",
      "value": 0.22186691498791333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01029",
        "aggregate_source_id": "aggregate-source-0097",
        "source_field": "/all_steps/inference_seconds/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01030",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01030",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/0/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01031",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "25.1213169515879",
      "value": 25.1213169515879,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01031",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/1/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01032",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.79999999999725",
      "value": 199.79999999999725,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01032",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/2/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01033",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "12723.524296082029",
      "value": 12723.524296082029,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01033",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/3/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01034",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01034",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/4/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01035",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "474.3696211129709",
      "value": 474.3696211129709,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01035",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/5/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01036",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01036",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/6/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01037",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "193.6866369832911",
      "value": 193.6866369832911,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01037",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/7/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01038",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01038",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/8/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01039",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "724.6637938974486",
      "value": 724.6637938974486,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01039",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/9/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01040",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01040",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/10/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01041",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4526.486241372053",
      "value": 4526.486241372053,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01041",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/11/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01042",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01042",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/12/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01043",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "588.2070436123427",
      "value": 588.2070436123427,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01043",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/13/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01044",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999721",
      "value": 199.99999999999721,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01044",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/14/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01045",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "latent_expected_reward_gap",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "2214.033129867404",
      "value": 2214.033129867404,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01045",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/15/mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01046",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "76.0",
      "value": 76.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01046",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/0/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01047",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "243.15263308533488",
      "value": 243.15263308533488,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01047",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/1/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01048",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "399.3333333333333",
      "value": 399.3333333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01048",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/2/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01049",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-7330.376736932822",
      "value": -7330.376736932822,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01049",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/3/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01050",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "75.9",
      "value": 75.9,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01050",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/4/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01051",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "133.65207743808114",
      "value": 133.65207743808114,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01051",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/5/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01052",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "401.1",
      "value": 401.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01052",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/6/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01053",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4803.737398226232",
      "value": 4803.737398226232,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01053",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/7/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01054",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "72.5",
      "value": 72.5,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01054",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/8/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01055",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-237.2638759840618",
      "value": -237.2638759840618,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01055",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/9/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01056",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "402.8",
      "value": 402.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01056",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/10/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01057",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "875.6746041235946",
      "value": 875.6746041235946,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01057",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/11/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01058",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "73.8",
      "value": 73.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01058",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/12/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01059",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "-188.55583630431218",
      "value": -188.55583630431218,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01059",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/13/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01060",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "400.8",
      "value": 400.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01060",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/14/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01061",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "autonomous_environment_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Each condition retains original reward family and horizon; never pooled across conditions.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "2316.702958645001",
      "value": 2316.702958645001,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01061",
        "aggregate_source_id": "aggregate-source-0098",
        "source_field": "/conditions/15/mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01102",
      "task": "hybrid_information_sampling",
      "model": "laya",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_test_loss_equal_subject_fold_mean",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "equal_subject_equal_original_fold_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "20 subjects, 160 original subject/fold units; 163135 total scored responses.",
      "n_raw": "20",
      "n": 20,
      "value_raw": "1.3924950981943565",
      "value": 1.3924950981943565,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01102",
        "aggregate_source_id": "aggregate-source-0099",
        "source_field": "/mean_test_loss_equal_subject_fold_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01103",
      "task": "hybrid_information_sampling",
      "model": "laya",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "1.3904522286236622",
      "value": 1.3904522286236622,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01103",
        "aggregate_source_id": "aggregate-source-0099",
        "source_field": "/pooled_action_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01104",
      "task": "hybrid_information_sampling",
      "model": "laya",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.05361049299244769",
      "value": 0.05361049299244769,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01104",
        "aggregate_source_id": "aggregate-source-0100",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01105",
      "task": "hybrid_information_sampling",
      "model": "laya",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.06993564000003971",
      "value": 0.06993564000003971,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01105",
        "aggregate_source_id": "aggregate-source-0100",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01106",
      "task": "hybrid_information_sampling",
      "model": "intern_decision_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_test_loss_equal_subject_fold_mean",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "equal_subject_equal_original_fold_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "20 subjects, 160 original subject/fold units; 163135 total scored responses.",
      "n_raw": "20",
      "n": 20,
      "value_raw": "1.1189720557560594",
      "value": 1.1189720557560594,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01106",
        "aggregate_source_id": "aggregate-source-0101",
        "source_field": "/mean_test_loss_equal_subject_fold_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01107",
      "task": "hybrid_information_sampling",
      "model": "intern_decision_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "1.116183035166427",
      "value": 1.116183035166427,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01107",
        "aggregate_source_id": "aggregate-source-0101",
        "source_field": "/pooled_action_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01108",
      "task": "hybrid_information_sampling",
      "model": "intern_decision_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.2606624700129032",
      "value": 0.2606624700129032,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01108",
        "aggregate_source_id": "aggregate-source-0102",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01109",
      "task": "hybrid_information_sampling",
      "model": "intern_decision_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.31918777601094916",
      "value": 0.31918777601094916,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01109",
        "aggregate_source_id": "aggregate-source-0102",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01110",
      "task": "hybrid_information_sampling",
      "model": "llama8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_test_loss_equal_subject_fold_mean",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "equal_subject_equal_original_fold_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "20 subjects, 160 original subject/fold units; 163135 total scored responses.",
      "n_raw": "20",
      "n": 20,
      "value_raw": "1.5310923339897151",
      "value": 1.5310923339897151,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01110",
        "aggregate_source_id": "aggregate-source-0103",
        "source_field": "/mean_test_loss_equal_subject_fold_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01111",
      "task": "hybrid_information_sampling",
      "model": "llama8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "1.5145696256504524",
      "value": 1.5145696256504524,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01111",
        "aggregate_source_id": "aggregate-source-0103",
        "source_field": "/pooled_action_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01112",
      "task": "hybrid_information_sampling",
      "model": "llama8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.08171006001066417",
      "value": 0.08171006001066417,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01112",
        "aggregate_source_id": "aggregate-source-0104",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01113",
      "task": "hybrid_information_sampling",
      "model": "llama8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.10521427700587083",
      "value": 0.10521427700587083,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01113",
        "aggregate_source_id": "aggregate-source-0104",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01114",
      "task": "hybrid_information_sampling",
      "model": "minitaur8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_test_loss_equal_subject_fold_mean",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "equal_subject_equal_original_fold_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "20 subjects, 160 original subject/fold units; 163135 total scored responses.",
      "n_raw": "20",
      "n": 20,
      "value_raw": "1.046602527088211",
      "value": 1.046602527088211,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01114",
        "aggregate_source_id": "aggregate-source-0105",
        "source_field": "/mean_test_loss_equal_subject_fold_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01115",
      "task": "hybrid_information_sampling",
      "model": "minitaur8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "1.0390852905973533",
      "value": 1.0390852905973533,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01115",
        "aggregate_source_id": "aggregate-source-0105",
        "source_field": "/pooled_action_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01116",
      "task": "hybrid_information_sampling",
      "model": "minitaur8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.07855786899744999",
      "value": 0.07855786899744999,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01116",
        "aggregate_source_id": "aggregate-source-0106",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01117",
      "task": "hybrid_information_sampling",
      "model": "minitaur8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.10161083399725612",
      "value": 0.10161083399725612,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01117",
        "aggregate_source_id": "aggregate-source-0106",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01118",
      "task": "hybrid_information_sampling",
      "model": "qwen35_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_test_loss_equal_subject_fold_mean",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "equal_subject_equal_original_fold_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "20 subjects, 160 original subject/fold units; 163135 total scored responses.",
      "n_raw": "20",
      "n": 20,
      "value_raw": "2.6443869861550935",
      "value": 2.6443869861550935,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01118",
        "aggregate_source_id": "aggregate-source-0107",
        "source_field": "/mean_test_loss_equal_subject_fold_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01119",
      "task": "hybrid_information_sampling",
      "model": "qwen35_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "2.6484975085989992",
      "value": 2.6484975085989992,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01119",
        "aggregate_source_id": "aggregate-source-0107",
        "source_field": "/pooled_action_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01120",
      "task": "hybrid_information_sampling",
      "model": "qwen35_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.2262541360105388",
      "value": 0.2262541360105388,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01120",
        "aggregate_source_id": "aggregate-source-0108",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01121",
      "task": "hybrid_information_sampling",
      "model": "qwen35_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.2673514629714191",
      "value": 0.2673514629714191,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01121",
        "aggregate_source_id": "aggregate-source-0108",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {}
    },
    {
      "row_id": "v01122",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "strict_sdk_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "1.716724620391506",
      "value": 1.716724620391506,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01122",
        "aggregate_source_id": "aggregate-source-0109",
        "source_field": "/strict_sdk_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01123",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "unrounded_mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "1.7167152841835003",
      "value": 1.7167152841835003,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01123",
        "aggregate_source_id": "aggregate-source-0109",
        "source_field": "/unrounded_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01124",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared GPU; measured elapsed time is not exclusive-device speed; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.10945578198879957",
      "value": 0.10945578198879957,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01124",
        "aggregate_source_id": "aggregate-source-0110",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01125",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared CPU/GPU; these measured times are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.13380020996555686",
      "value": 0.13380020996555686,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01125",
        "aggregate_source_id": "aggregate-source-0111",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01126",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared GPU; measured elapsed time is not exclusive-device speed; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.11874378751963377",
      "value": 0.11874378751963377,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01126",
        "aggregate_source_id": "aggregate-source-0110",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01127",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared CPU/GPU; these measured times are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.14777351822704077",
      "value": 0.14777351822704077,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01127",
        "aggregate_source_id": "aggregate-source-0111",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01128",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "mean_reveals",
      "metric_variant": "mean_reveals",
      "scope": "all",
      "target": "mean_reveals",
      "unit": "revealed_nodes_per_generated_trial",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01128",
        "aggregate_source_id": "aggregate-source-0112",
        "source_field": "/mean_reveals"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01129",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "all",
      "target": "best_path_expected_value_at_STOP",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Cost=0, expected path value; not realized route payoff.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01129",
        "aggregate_source_id": "aggregate-source-0112",
        "source_field": "/mean_terminal_expected_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01130",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "2.1092053428276376",
      "value": 2.1092053428276376,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01130",
        "aggregate_source_id": "aggregate-source-0113",
        "source_field": "/mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01131",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.18415803695097566",
      "value": 0.18415803695097566,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01131",
        "aggregate_source_id": "aggregate-source-0114",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01132",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.20495075098006055",
      "value": 0.20495075098006055,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01132",
        "aggregate_source_id": "aggregate-source-0114",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01133",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "2.7433772285300697",
      "value": 2.7433772285300697,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01133",
        "aggregate_source_id": "aggregate-source-0115",
        "source_field": "/mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01134",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared CPU/GPU; these measured times are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "47500",
      "n": 47500,
      "value_raw": "0.15857447078451514",
      "value": 0.15857447078451514,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01134",
        "aggregate_source_id": "aggregate-source-0116",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01135",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.10189179982990026",
      "value": 0.10189179982990026,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01135",
        "aggregate_source_id": "aggregate-source-0117",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01136",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared CPU/GPU; these measured times are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "47500",
      "n": 47500,
      "value_raw": "0.19356313906610012",
      "value": 0.19356313906610012,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01136",
        "aggregate_source_id": "aggregate-source-0116",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01137",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.10878027603030205",
      "value": 0.10878027603030205,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01137",
        "aggregate_source_id": "aggregate-source-0117",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01138",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "mean_reveals",
      "metric_variant": "mean_reveals",
      "scope": "all",
      "target": "mean_reveals",
      "unit": "revealed_nodes_per_generated_trial",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "1.0",
      "value": 1.0,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01138",
        "aggregate_source_id": "aggregate-source-0118",
        "source_field": "/mean_reveals"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01139",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "all",
      "target": "best_path_expected_value_at_STOP",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Cost=0, expected path value; not realized route payoff.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "3.7658947368421054",
      "value": 3.7658947368421054,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01139",
        "aggregate_source_id": "aggregate-source-0118",
        "source_field": "/mean_terminal_expected_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01140",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "2.1248951951577566",
      "value": 2.1248951951577566,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01140",
        "aggregate_source_id": "aggregate-source-0119",
        "source_field": "/mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01141",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared CPU/GPU; these measured times are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.156759868375957",
      "value": 0.156759868375957,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01141",
        "aggregate_source_id": "aggregate-source-0120",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01142",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.10144483856856823",
      "value": 0.10144483856856823,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01142",
        "aggregate_source_id": "aggregate-source-0121",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01143",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared CPU/GPU; these measured times are not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.15905303694307804",
      "value": 0.15905303694307804,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01143",
        "aggregate_source_id": "aggregate-source-0120",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01144",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.10736708529293537",
      "value": 0.10736708529293537,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01144",
        "aggregate_source_id": "aggregate-source-0121",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01145",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "mean_reveals",
      "metric_variant": "mean_reveals",
      "scope": "all",
      "target": "mean_reveals",
      "unit": "revealed_nodes_per_generated_trial",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01145",
        "aggregate_source_id": "aggregate-source-0122",
        "source_field": "/mean_reveals"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01146",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_metamdp_generative_23750",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "all",
      "target": "best_path_expected_value_at_STOP",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Cost=0, expected path value; not realized route payoff.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01146",
        "aggregate_source_id": "aggregate-source-0122",
        "source_field": "/mean_terminal_expected_reward"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01147",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "2.4637485369829424",
      "value": 2.4637485369829424,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01147",
        "aggregate_source_id": "aggregate-source-0123",
        "source_field": "/mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01148",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.15710287203546613",
      "value": 0.15710287203546613,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01148",
        "aggregate_source_id": "aggregate-source-0124",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01149",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.18889479903737083",
      "value": 0.18889479903737083,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01149",
        "aggregate_source_id": "aggregate-source-0124",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01172",
      "task": "bart",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "24503",
      "n": 24503,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01172",
        "aggregate_source_id": "aggregate-source-0125",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01173",
      "task": "bart",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "24503",
      "n": 24503,
      "value_raw": "0.2896392050315626",
      "value": 0.2896392050315626,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01173",
        "aggregate_source_id": "aggregate-source-0125",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01174",
      "task": "bart",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "24503",
      "n": 24503,
      "value_raw": "0.6280707357800562",
      "value": 0.6280707357800562,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01174",
        "aggregate_source_id": "aggregate-source-0125",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01175",
      "task": "bart",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "900",
      "n": 900,
      "value_raw": "0.0044444444444444444",
      "value": 0.0044444444444444444,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01175",
        "aggregate_source_id": "aggregate-source-0125",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01176",
      "task": "bart",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "11997",
      "n": 11997,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01176",
        "aggregate_source_id": "aggregate-source-0126",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01177",
      "task": "bart",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "11997",
      "n": 11997,
      "value_raw": "1.056215652031824",
      "value": 1.056215652031824,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01177",
        "aggregate_source_id": "aggregate-source-0126",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01178",
      "task": "bart",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "11997",
      "n": 11997,
      "value_raw": "1.8753830455942078",
      "value": 1.8753830455942078,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01178",
        "aggregate_source_id": "aggregate-source-0126",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01179",
      "task": "bart",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "900",
      "n": 900,
      "value_raw": "0.7422222222222222",
      "value": 0.7422222222222222,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01179",
        "aggregate_source_id": "aggregate-source-0126",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01180",
      "task": "bart",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1166",
      "n": 1166,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01180",
        "aggregate_source_id": "aggregate-source-0127",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01181",
      "task": "bart",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "1166",
      "n": 1166,
      "value_raw": "0.2700382679759059",
      "value": 0.2700382679759059,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01181",
        "aggregate_source_id": "aggregate-source-0127",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01182",
      "task": "bart",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "1166",
      "n": 1166,
      "value_raw": "0.46873251275974326",
      "value": 0.46873251275974326,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01182",
        "aggregate_source_id": "aggregate-source-0127",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01183",
      "task": "bart",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "900",
      "n": 900,
      "value_raw": "0.2811111111111111",
      "value": 0.2811111111111111,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01183",
        "aggregate_source_id": "aggregate-source-0127",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01184",
      "task": "bart",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "25031",
      "n": 25031,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01184",
        "aggregate_source_id": "aggregate-source-0128",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01185",
      "task": "bart",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "25031",
      "n": 25031,
      "value_raw": "0.4628503500134684",
      "value": 0.4628503500134684,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01185",
        "aggregate_source_id": "aggregate-source-0128",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01186",
      "task": "bart",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "25031",
      "n": 25031,
      "value_raw": "0.8132446280214936",
      "value": 0.8132446280214936,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01186",
        "aggregate_source_id": "aggregate-source-0128",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01187",
      "task": "bart",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "900",
      "n": 900,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01187",
        "aggregate_source_id": "aggregate-source-0128",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01188",
      "task": "bart",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "25031",
      "n": 25031,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01188",
        "aggregate_source_id": "aggregate-source-0129",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01189",
      "task": "bart",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "25031",
      "n": 25031,
      "value_raw": "0.4246633579605259",
      "value": 0.4246633579605259,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01189",
        "aggregate_source_id": "aggregate-source-0129",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01190",
      "task": "bart",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "25031",
      "n": 25031,
      "value_raw": "0.7090217420191038",
      "value": 0.7090217420191038,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01190",
        "aggregate_source_id": "aggregate-source-0129",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01191",
      "task": "bart",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "900",
      "n": 900,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01191",
        "aggregate_source_id": "aggregate-source-0129",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01232",
      "task": "temporal_discounting",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01232",
        "aggregate_source_id": "aggregate-source-0130",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cpu"
      }
    },
    {
      "row_id": "v01233",
      "task": "temporal_discounting",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.2349878679960966",
      "value": 0.2349878679960966,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01233",
        "aggregate_source_id": "aggregate-source-0130",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cpu"
      }
    },
    {
      "row_id": "v01234",
      "task": "temporal_discounting",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.260386606794782",
      "value": 0.260386606794782,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01234",
        "aggregate_source_id": "aggregate-source-0130",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cpu"
      }
    },
    {
      "row_id": "v01235",
      "task": "temporal_discounting",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01235",
        "aggregate_source_id": "aggregate-source-0131",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01236",
      "task": "temporal_discounting",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.41605792799964547",
      "value": 0.41605792799964547,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01236",
        "aggregate_source_id": "aggregate-source-0131",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01237",
      "task": "temporal_discounting",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.7419103731866924",
      "value": 0.7419103731866924,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01237",
        "aggregate_source_id": "aggregate-source-0131",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01238",
      "task": "temporal_discounting",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01238",
        "aggregate_source_id": "aggregate-source-0132",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01239",
      "task": "temporal_discounting",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.15795643901219591",
      "value": 0.15795643901219591,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01239",
        "aggregate_source_id": "aggregate-source-0132",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01240",
      "task": "temporal_discounting",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.39388751122169186",
      "value": 0.39388751122169186,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01240",
        "aggregate_source_id": "aggregate-source-0132",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01241",
      "task": "temporal_discounting",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "13",
      "n": 13,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01241",
        "aggregate_source_id": "aggregate-source-0133",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01242",
      "task": "temporal_discounting",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "13",
      "n": 13,
      "value_raw": "0.1304296599701047",
      "value": 0.1304296599701047,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01242",
        "aggregate_source_id": "aggregate-source-0133",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01243",
      "task": "temporal_discounting",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "13",
      "n": 13,
      "value_raw": "0.2796050904202271",
      "value": 0.2796050904202271,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01243",
        "aggregate_source_id": "aggregate-source-0133",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01244",
      "task": "temporal_discounting",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "13",
      "n": 13,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01244",
        "aggregate_source_id": "aggregate-source-0134",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01245",
      "task": "temporal_discounting",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "13",
      "n": 13,
      "value_raw": "0.1323386179865338",
      "value": 0.1323386179865338,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01245",
        "aggregate_source_id": "aggregate-source-0134",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01246",
      "task": "temporal_discounting",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared host/device; engineering latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "13",
      "n": 13,
      "value_raw": "3.1692247552215016",
      "value": 3.1692247552215016,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01246",
        "aggregate_source_id": "aggregate-source-0134",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01247",
      "task": "two_step",
      "model": "laya",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.17773523880168796",
      "value": 0.17773523880168796,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01247",
        "aggregate_source_id": "aggregate-source-0135",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01248",
      "task": "two_step",
      "model": "laya",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.24554519075900316",
      "value": 0.24554519075900316,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01248",
        "aggregate_source_id": "aggregate-source-0135",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01249",
      "task": "two_step",
      "model": "laya",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2000",
      "n": 2000,
      "value_raw": "0.4965",
      "value": 0.4965,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01249",
        "aggregate_source_id": "aggregate-source-0136",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01250",
      "task": "two_step",
      "model": "laya",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "stay_reward_transition_interaction",
      "metric_variant": "stay_reward_common_interaction_original_OLS",
      "scope": "all",
      "target": "stay_regressed_on_reward_times_common_transition",
      "unit": "coefficient",
      "aggregation": "original_OLS",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1900",
      "n": 1900,
      "value_raw": "0.018631059452397222",
      "value": 0.018631059452397222,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01250",
        "aggregate_source_id": "aggregate-source-0136",
        "source_field": "/stay_reward_common_interaction_original_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01251",
      "task": "two_step",
      "model": "intern_decision_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.2328731970046647",
      "value": 0.2328731970046647,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01251",
        "aggregate_source_id": "aggregate-source-0137",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01252",
      "task": "two_step",
      "model": "intern_decision_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.31659241899615154",
      "value": 0.31659241899615154,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01252",
        "aggregate_source_id": "aggregate-source-0137",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01253",
      "task": "two_step",
      "model": "intern_decision_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2000",
      "n": 2000,
      "value_raw": "0.523",
      "value": 0.523,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01253",
        "aggregate_source_id": "aggregate-source-0138",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01254",
      "task": "two_step",
      "model": "intern_decision_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "stay_reward_transition_interaction",
      "metric_variant": "stay_reward_common_interaction_original_OLS",
      "scope": "all",
      "target": "stay_regressed_on_reward_times_common_transition",
      "unit": "coefficient",
      "aggregation": "original_OLS",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1900",
      "n": 1900,
      "value_raw": "0.16385210396963798",
      "value": 0.16385210396963798,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01254",
        "aggregate_source_id": "aggregate-source-0138",
        "source_field": "/stay_reward_common_interaction_original_OLS"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01255",
      "task": "two_step",
      "model": "llama8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.105230912566185",
      "value": 0.105230912566185,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01255",
        "aggregate_source_id": "aggregate-source-0139",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01256",
      "task": "two_step",
      "model": "llama8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.15376584418118",
      "value": 0.15376584418118,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01256",
        "aggregate_source_id": "aggregate-source-0139",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01257",
      "task": "two_step",
      "model": "llama8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2000",
      "n": 2000,
      "value_raw": "0.5035",
      "value": 0.5035,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01257",
        "aggregate_source_id": "aggregate-source-0140",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01258",
      "task": "two_step",
      "model": "llama8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "stay_reward_transition_interaction",
      "metric_variant": "stay_reward_common_interaction_original_OLS",
      "scope": "all",
      "target": "stay_regressed_on_reward_times_common_transition",
      "unit": "coefficient",
      "aggregation": "original_OLS",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1900",
      "n": 1900,
      "value_raw": "7.28583859910259e-17",
      "value": 7.28583859910259e-17,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01258",
        "aggregate_source_id": "aggregate-source-0140",
        "source_field": "/stay_reward_common_interaction_original_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01259",
      "task": "two_step",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.161596923135221",
      "value": 0.161596923135221,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01259",
        "aggregate_source_id": "aggregate-source-0141",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01260",
      "task": "two_step",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.22446109540760517",
      "value": 0.22446109540760517,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01260",
        "aggregate_source_id": "aggregate-source-0141",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01261",
      "task": "two_step",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2000",
      "n": 2000,
      "value_raw": "0.509",
      "value": 0.509,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01261",
        "aggregate_source_id": "aggregate-source-0142",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01262",
      "task": "two_step",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "stay_reward_transition_interaction",
      "metric_variant": "stay_reward_common_interaction_original_OLS",
      "scope": "all",
      "target": "stay_regressed_on_reward_times_common_transition",
      "unit": "coefficient",
      "aggregation": "original_OLS",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1900",
      "n": 1900,
      "value_raw": "0.006729962573554079",
      "value": 0.006729962573554079,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01262",
        "aggregate_source_id": "aggregate-source-0142",
        "source_field": "/stay_reward_common_interaction_original_OLS"
      },
      "runtime_context": {
        "device": "cuda:0"
      }
    },
    {
      "row_id": "v01263",
      "task": "two_step",
      "model": "qwen35_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.1948312220047228",
      "value": 0.1948312220047228,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01263",
        "aggregate_source_id": "aggregate-source-0143",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01264",
      "task": "two_step",
      "model": "qwen35_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared machine; elapsed time is not exclusive-device performance.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.2767833399702795",
      "value": 0.2767833399702795,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01264",
        "aggregate_source_id": "aggregate-source-0143",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01265",
      "task": "two_step",
      "model": "qwen35_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "2000",
      "n": 2000,
      "value_raw": "0.519",
      "value": 0.519,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01265",
        "aggregate_source_id": "aggregate-source-0144",
        "source_field": "/mean_reward"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01266",
      "task": "two_step",
      "model": "qwen35_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "stay_reward_transition_interaction",
      "metric_variant": "stay_reward_common_interaction_original_OLS",
      "scope": "all",
      "target": "stay_regressed_on_reward_times_common_transition",
      "unit": "coefficient",
      "aggregation": "original_OLS",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "1900",
      "n": 1900,
      "value_raw": "0.09321703918667174",
      "value": 0.09321703918667174,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01266",
        "aggregate_source_id": "aggregate-source-0144",
        "source_field": "/stay_reward_common_interaction_original_OLS"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01300",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "ToMi/test",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.39522856189522854",
      "value": 0.39522856189522854,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01300",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/by_experiment/ToMi~1test/original_answer_exact_match_mean"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01301",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.39522856189522854",
      "value": 0.39522856189522854,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01301",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/original_answer_exact_match_mean"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01302",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.39522856189522854",
      "value": 0.39522856189522854,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01302",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/27/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01303",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.7937754770003336",
      "value": 0.7937754770003336,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01303",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/27/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01304",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "ToMi/test",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01304",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/by_experiment/ToMi~1test/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01305",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01305",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01306",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.12557946200831793",
      "value": 0.12557946200831793,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01306",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/by_experiment/ToMi~1test/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01307",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.12557946200831793",
      "value": 0.12557946200831793,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01307",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01308",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.13778675159846898",
      "value": 0.13778675159846898,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01308",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/by_experiment/ToMi~1test/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01309",
      "task": "false_belief_story_qa",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.13778675159846898",
      "value": 0.13778675159846898,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01309",
        "aggregate_source_id": "aggregate-source-0145",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01310",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "ToMi/test",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.824657991324658",
      "value": 0.824657991324658,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01310",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/by_experiment/ToMi~1test/original_answer_exact_match_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01311",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.824657991324658",
      "value": 0.824657991324658,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01311",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/original_answer_exact_match_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01312",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.824657991324658",
      "value": 0.824657991324658,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01312",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/25/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01313",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.2282139001499705",
      "value": 0.2282139001499705,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01313",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/25/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01314",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "ToMi/test",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01314",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/by_experiment/ToMi~1test/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01315",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01315",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01316",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.1418400394904893",
      "value": 0.1418400394904893,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01316",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/by_experiment/ToMi~1test/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01317",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.1418400394904893",
      "value": 0.1418400394904893,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01317",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01318",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.15404128497757483",
      "value": 0.15404128497757483,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01318",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/by_experiment/ToMi~1test/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01319",
      "task": "false_belief_story_qa",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.15404128497757483",
      "value": 0.15404128497757483,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01319",
        "aggregate_source_id": "aggregate-source-0146",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01320",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "ToMi/test",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.8284951618284951",
      "value": 0.8284951618284951,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01320",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/by_experiment/ToMi~1test/original_answer_exact_match_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01321",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.8284951618284951",
      "value": 0.8284951618284951,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01321",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/original_answer_exact_match_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01322",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.8284951618284951",
      "value": 0.8284951618284951,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01322",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/26/actual_objective_accuracy_check"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01323",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.2910312916397705",
      "value": 0.2910312916397705,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01323",
        "aggregate_source_id": "aggregate-source-0013",
        "source_field": "/runs/26/objective_answer_brier_sum_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01324",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "ToMi/test",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01324",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/by_experiment/ToMi~1test/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01325",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01325",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01326",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.12260539550334215",
      "value": 0.12260539550334215,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01326",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/by_experiment/ToMi~1test/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01327",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.12260539550334215",
      "value": 0.12260539550334215,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01327",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01328",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.1374473367148312",
      "value": 0.1374473367148312,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01328",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/by_experiment/ToMi~1test/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01329",
      "task": "false_belief_story_qa",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.1374473367148312",
      "value": 0.1374473367148312,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01329",
        "aggregate_source_id": "aggregate-source-0147",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01330",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "128",
      "n": 128,
      "value_raw": "1.1920789604071451",
      "value": 1.1920789604071451,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01330",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01331",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01331",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01332",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp2",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01332",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01333",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01333",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01334",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.13472853248822503",
      "value": 0.13472853248822503,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01334",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01335",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.16715810948517174",
      "value": 0.16715810948517174,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01335",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01336",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.16192006997880526",
      "value": 0.16192006997880526,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01336",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01337",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.14774538956116884",
      "value": 0.14774538956116884,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01337",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01338",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.1851616874919273",
      "value": 0.1851616874919273,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01338",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01339",
      "task": "politeness_pragmatics",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.18513046930893323",
      "value": 0.18513046930893323,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01339",
        "aggregate_source_id": "aggregate-source-0148",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01340",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.7988451649484437",
      "value": 0.7988451649484437,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01340",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01341",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01341",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01342",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp2",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01342",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01343",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01343",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01344",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.15061439652345143",
      "value": 0.15061439652345143,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01344",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01345",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.18931273248745129",
      "value": 0.18931273248745129,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01345",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01346",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.18783821249962784",
      "value": 0.18783821249962784,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01346",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01347",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.1608392710943008",
      "value": 0.1608392710943008,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01347",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01348",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.19819651622674428",
      "value": 0.19819651622674428,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01348",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01349",
      "task": "politeness_pragmatics",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.19814863513747696",
      "value": 0.19814863513747696,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01349",
        "aggregate_source_id": "aggregate-source-0149",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01350",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.9998605394748019",
      "value": 0.9998605394748019,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01350",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01351",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01351",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01352",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp2",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01352",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01353",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01353",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01354",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.13795868199667893",
      "value": 0.13795868199667893,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01354",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01355",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.15114910050760955",
      "value": 0.15114910050760955,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01355",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01356",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.14957712253089994",
      "value": 0.14957712253089994,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01356",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01357",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.1501928185520228",
      "value": 0.1501928185520228,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01357",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01358",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.1610503910051193",
      "value": 0.1610503910051193,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01358",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01359",
      "task": "politeness_pragmatics",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.16100263762346004",
      "value": 0.16100263762346004,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01359",
        "aggregate_source_id": "aggregate-source-0150",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01360",
      "task": "pragmatic_language_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "1.5915153074312387",
      "value": 1.5915153074312387,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01360",
        "aggregate_source_id": "aggregate-source-0151",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01361",
      "task": "pragmatic_language_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Hu2023Fine/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01361",
        "aggregate_source_id": "aggregate-source-0151",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01362",
      "task": "pragmatic_language_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01362",
        "aggregate_source_id": "aggregate-source-0151",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01363",
      "task": "pragmatic_language_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.16639661602675915",
      "value": 0.16639661602675915,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01363",
        "aggregate_source_id": "aggregate-source-0151",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01364",
      "task": "pragmatic_language_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.16639661602675915",
      "value": 0.16639661602675915,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01364",
        "aggregate_source_id": "aggregate-source-0151",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01365",
      "task": "pragmatic_language_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.193193933472503",
      "value": 0.193193933472503,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01365",
        "aggregate_source_id": "aggregate-source-0151",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01366",
      "task": "pragmatic_language_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.193193933472503",
      "value": 0.193193933472503,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01366",
        "aggregate_source_id": "aggregate-source-0151",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01367",
      "task": "pragmatic_language_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.8137968669470528",
      "value": 0.8137968669470528,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01367",
        "aggregate_source_id": "aggregate-source-0152",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01368",
      "task": "pragmatic_language_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Hu2023Fine/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01368",
        "aggregate_source_id": "aggregate-source-0152",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01369",
      "task": "pragmatic_language_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01369",
        "aggregate_source_id": "aggregate-source-0152",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01370",
      "task": "pragmatic_language_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.17954748700140044",
      "value": 0.17954748700140044,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01370",
        "aggregate_source_id": "aggregate-source-0152",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01371",
      "task": "pragmatic_language_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.17954748700140044",
      "value": 0.17954748700140044,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01371",
        "aggregate_source_id": "aggregate-source-0152",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01372",
      "task": "pragmatic_language_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.2042279974237317",
      "value": 0.2042279974237317,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01372",
        "aggregate_source_id": "aggregate-source-0152",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01373",
      "task": "pragmatic_language_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.2042279974237317",
      "value": 0.2042279974237317,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01373",
        "aggregate_source_id": "aggregate-source-0152",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01374",
      "task": "pragmatic_language_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "1.2395484997252715",
      "value": 1.2395484997252715,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01374",
        "aggregate_source_id": "aggregate-source-0153",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01375",
      "task": "pragmatic_language_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Hu2023Fine/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01375",
        "aggregate_source_id": "aggregate-source-0153",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01376",
      "task": "pragmatic_language_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01376",
        "aggregate_source_id": "aggregate-source-0153",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01377",
      "task": "pragmatic_language_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.15570575397578068",
      "value": 0.15570575397578068,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01377",
        "aggregate_source_id": "aggregate-source-0153",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01378",
      "task": "pragmatic_language_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.15570575397578068",
      "value": 0.15570575397578068,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01378",
        "aggregate_source_id": "aggregate-source-0153",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01379",
      "task": "pragmatic_language_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.19608118026226293",
      "value": 0.19608118026226293,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01379",
        "aggregate_source_id": "aggregate-source-0153",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01380",
      "task": "pragmatic_language_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.19608118026226293",
      "value": 0.19608118026226293,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01380",
        "aggregate_source_id": "aggregate-source-0153",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01381",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01381",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01382",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01382",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01383",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01383",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01384",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.34629542849143036",
      "value": 0.34629542849143036,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01384",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01385",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.1487802899791859",
      "value": 0.1487802899791859,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01385",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01386",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.3447070420079399",
      "value": 0.3447070420079399,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01386",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01387",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.3609119328728411",
      "value": 0.3609119328728411,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01387",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01388",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.15733067469263914",
      "value": 0.15733067469263914,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01388",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01389",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.3608691864006687",
      "value": 0.3608691864006687,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01389",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01390",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "300",
      "n": 300,
      "value_raw": "30.486166666666666",
      "value": 30.486166666666666,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01390",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01391",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "30",
      "n": 30,
      "value_raw": "34.61366666666667",
      "value": 34.61366666666667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01391",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01392",
      "task": "nonliteral_number_interpretation",
      "model": "laya",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "330",
      "n": 330,
      "value_raw": "30.86139393939394",
      "value": 30.86139393939394,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01392",
        "aggregate_source_id": "aggregate-source-0154",
        "source_field": "/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:1",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01393",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01393",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01394",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01394",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01395",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01395",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01396",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.3720071234856732",
      "value": 0.3720071234856732,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01396",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01397",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.18469477849430405",
      "value": 0.18469477849430405,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01397",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01398",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.3705723340099212",
      "value": 0.3705723340099212,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01398",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01399",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.3849554048880236",
      "value": 0.3849554048880236,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01399",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01400",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.19305839323496912",
      "value": 0.19305839323496912,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01400",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01401",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.38436202526499985",
      "value": 0.38436202526499985,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01401",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01402",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "300",
      "n": 300,
      "value_raw": "17.10159246551691",
      "value": 17.10159246551691,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01402",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01403",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "30",
      "n": 30,
      "value_raw": "14.548255644301111",
      "value": 14.548255644301111,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01403",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01404",
      "task": "nonliteral_number_interpretation",
      "model": "intern_decision_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "330",
      "n": 330,
      "value_raw": "16.869470936315473",
      "value": 16.869470936315473,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01404",
        "aggregate_source_id": "aggregate-source-0155",
        "source_field": "/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01405",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01405",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01406",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01406",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01407",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01407",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01408",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.48525632251403295",
      "value": 0.48525632251403295,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01408",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01409",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.3792495310190134",
      "value": 0.3792495310190134,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01409",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01410",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.483236403495539",
      "value": 0.483236403495539,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01410",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01411",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.6098623752739514",
      "value": 0.6098623752739514,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01411",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01412",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.3874916332249995",
      "value": 0.3874916332249995,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01412",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01413",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.6073462347179884",
      "value": 0.6073462347179884,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01413",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01414",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "300",
      "n": 300,
      "value_raw": "30.47833333333333",
      "value": 30.47833333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01414",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01415",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "30",
      "n": 30,
      "value_raw": "22.473666666666666",
      "value": 22.473666666666666,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01415",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01416",
      "task": "nonliteral_number_interpretation",
      "model": "qwen35_4b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "330",
      "n": 330,
      "value_raw": "29.750636363636364",
      "value": 29.750636363636364,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01416",
        "aggregate_source_id": "aggregate-source-0156",
        "source_field": "/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "cuda:0",
        "batch_size": 1
      }
    },
    {
      "row_id": "v01417",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.23215209518882565",
      "value": 0.23215209518882565,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01417",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/all/objective_accuracy"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01418",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.19860681114551085",
      "value": 0.19860681114551085,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01418",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01419",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.19860681114551085",
      "value": 0.19860681114551085,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01419",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01420",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.23521027377928308",
      "value": 0.23521027377928308,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01420",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01421",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.23521027377928308",
      "value": 0.23521027377928308,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01421",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01422",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum_k(p_k-onehot_k)^2)",
      "scope": "grammar_and_logic_objective_answer_rows",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.8220485653643614",
      "value": 0.8220485653643614,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01422",
        "aggregate_source_id": "aggregate-source-0158",
        "source_field": "/objective_answer_rows/mean(sum_k(p_k-onehot_k)^2)"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01423",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01423",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01424",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "9044",
      "n": 9044,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01424",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01425",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "5.4436776719029325",
      "value": 5.4436776719029325,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01425",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01426",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "1.6846941743073591",
      "value": 1.6846941743073591,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01426",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01427",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.943614159422056",
      "value": 2.943614159422056,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01427",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01428",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01428",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01429",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01429",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01430",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "99204",
      "n": 99204,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01430",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01431",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01431",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01432",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "1.6771444637542923",
      "value": 1.6771444637542923,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01432",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01433",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01433",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01434",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01434",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01435",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01435",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01436",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_response)/expected_responses",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_responses",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01436",
        "aggregate_source_id": "aggregate-source-0158",
        "source_field": "/all_records/sum(invalid_response)~1expected_responses"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01437",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.6078426890308037",
      "value": 0.6078426890308037,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01437",
        "aggregate_source_id": "aggregate-source-0158",
        "source_field": "/rows/latency_p50"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01438",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "1.3098869749810547",
      "value": 1.3098869749810547,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01438",
        "aggregate_source_id": "aggregate-source-0158",
        "source_field": "/rows/latency_p95"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01439",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "5.543343653250774",
      "value": 5.543343653250774,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01439",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01440",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "3.6191950464396285",
      "value": 3.6191950464396285,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01440",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01441",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.086687306501548",
      "value": 2.086687306501548,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01441",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01442",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "646",
      "n": 646,
      "value_raw": "39.30185758513932",
      "value": 39.30185758513932,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01442",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01443",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "7.012559977420265",
      "value": 7.012559977420265,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01443",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01444",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "4.422664408693198",
      "value": 4.422664408693198,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01444",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01445",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.5499576629974596",
      "value": 2.5499576629974596,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01445",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "v01446",
      "task": "performance_self_assessment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "37.63223257126729",
      "value": 37.63223257126729,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "v01446",
        "aggregate_source_id": "aggregate-source-0157",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {
        "device": "cuda:1"
      }
    },
    {
      "row_id": "new_rlwm_llama8b_full_1",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.417455727542477",
      "value": 0.417455727542477,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_1",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_2",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.417455727542477",
      "value": 0.417455727542477,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_2",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_3",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4178526204380942",
      "value": 0.4178526204380942,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_3",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/human_response_agreement"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_4",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4178526204380942",
      "value": 0.4178526204380942,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_4",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_5",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "same_original_human_history_response_no_official_heldout_split",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "2.1703925272585733",
      "value": 2.1703925272585733,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_5",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/typed_choice_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_6",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "same_original_human_history_response_no_official_heldout_split",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "2.1703925272585733",
      "value": 2.1703925272585733,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_6",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_7",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_7",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_8",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_8",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_9",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.09828998500597663",
      "value": 0.09828998500597663,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_9",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_rlwm_llama8b_full_10",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.21387456799857318",
      "value": 0.21387456799857318,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_rlwm_llama8b_full_10",
        "aggregate_source_id": "aggregate-source-0159",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_11",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "1.5838268399344961",
      "value": 1.5838268399344961,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_11",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_12",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_12",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_13",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp2",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_13",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_14",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_14",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_15",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.07696449250215665",
      "value": 0.07696449250215665,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_15",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_16",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.10099679400445893",
      "value": 0.10099679400445893,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_16",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_17",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.09320120001211762",
      "value": 0.09320120001211762,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_17",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_18",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.09484813244780525",
      "value": 0.09484813244780525,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_18",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_19",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.10618633649573894",
      "value": 0.10618633649573894,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_19",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_politeness_llama8b_full_20",
      "task": "politeness_pragmatics",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.10613482029584702",
      "value": 0.10613482029584702,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_politeness_llama8b_full_20",
        "aggregate_source_id": "aggregate-source-0160",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_pragmatic_llama8b_full_21",
      "task": "pragmatic_language_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "1.8514305638020085",
      "value": 1.8514305638020085,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_pragmatic_llama8b_full_21",
        "aggregate_source_id": "aggregate-source-0161",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_pragmatic_llama8b_full_22",
      "task": "pragmatic_language_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Hu2023Fine/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_pragmatic_llama8b_full_22",
        "aggregate_source_id": "aggregate-source-0161",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_pragmatic_llama8b_full_23",
      "task": "pragmatic_language_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_pragmatic_llama8b_full_23",
        "aggregate_source_id": "aggregate-source-0161",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_pragmatic_llama8b_full_24",
      "task": "pragmatic_language_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.08869643800426275",
      "value": 0.08869643800426275,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_pragmatic_llama8b_full_24",
        "aggregate_source_id": "aggregate-source-0161",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_pragmatic_llama8b_full_25",
      "task": "pragmatic_language_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.08869643800426275",
      "value": 0.08869643800426275,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_pragmatic_llama8b_full_25",
        "aggregate_source_id": "aggregate-source-0161",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_pragmatic_llama8b_full_26",
      "task": "pragmatic_language_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.13343600465450436",
      "value": 0.13343600465450436,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_pragmatic_llama8b_full_26",
        "aggregate_source_id": "aggregate-source-0161",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_pragmatic_llama8b_full_27",
      "task": "pragmatic_language_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.13343600465450436",
      "value": 0.13343600465450436,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_pragmatic_llama8b_full_27",
        "aggregate_source_id": "aggregate-source-0161",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_28",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_28",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_29",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_29",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_30",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_30",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_31",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.24352910050947685",
      "value": 0.24352910050947685,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_31",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_32",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.12118918950727675",
      "value": 0.12118918950727675,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_32",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_33",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.24103424500208348",
      "value": 0.24103424500208348,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_33",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_34",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.3558757634367794",
      "value": 0.3558757634367794,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_34",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_35",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.13936845189746236",
      "value": 0.13936845189746236,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_35",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_36",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.3553717409624369",
      "value": 0.3553717409624369,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_36",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_37",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "26.709533333333333",
      "value": 26.709533333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_37",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_38",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "8.950333333333335",
      "value": 8.950333333333335,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_38",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_nonliteral_llama8b_full_39",
      "task": "nonliteral_number_interpretation",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "target": "public_human_mean_slider_judgment",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "25.095060606060606",
      "value": 25.095060606060606,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_nonliteral_llama8b_full_39",
        "aggregate_source_id": "aggregate-source-0162",
        "source_field": "/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_40",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "ToMi/test",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.8183183183183184",
      "value": 0.8183183183183184,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_40",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/by_experiment/ToMi~1test/original_answer_exact_match_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_41",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.8183183183183184",
      "value": 0.8183183183183184,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_41",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/original_answer_exact_match_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_42",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "ToMi/test",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_42",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/by_experiment/ToMi~1test/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_43",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_43",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_44",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.08526789949974045",
      "value": 0.08526789949974045,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_44",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/by_experiment/ToMi~1test/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_45",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.08526789949974045",
      "value": 0.08526789949974045,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_45",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_46",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.08986875229602447",
      "value": 0.08986875229602447,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_46",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/by_experiment/ToMi~1test/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_psych_social_tomi_llama8b_full_47",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Newly completed source; original metric definition and scope retained. No score recalculation; shared-device timing remains separate.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.08986875229602447",
      "value": 0.08986875229602447,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_psych_social_tomi_llama8b_full_47",
        "aggregate_source_id": "aggregate-source-0163",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_1",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.189187790998448",
      "value": 0.189187790998448,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_1",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/all/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_2",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.7001547987616099",
      "value": 0.7001547987616099,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_2",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_3",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.7001547987616099",
      "value": 0.7001547987616099,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_3",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_4",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.14260513688964155",
      "value": 0.14260513688964155,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_4",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_5",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.14260513688964155",
      "value": 0.14260513688964155,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_5",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_6",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "1.8851206979439843",
      "value": 1.8851206979439843,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_6",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_7",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "9044",
      "n": 9044,
      "value_raw": "1.6204117613224445",
      "value": 1.6204117613224445,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_7",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_8",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.594243082954053",
      "value": 2.594243082954053,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_8",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_9",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "1.2642765432512604",
      "value": 1.2642765432512604,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_9",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_10",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.951911625847153",
      "value": 1.951911625847153,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_10",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_11",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.7579043065700388",
      "value": 1.7579043065700388,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_11",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_12",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "3.738940210630389",
      "value": 3.738940210630389,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_12",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_13",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "99204",
      "n": 99204,
      "value_raw": "1.9092530678363373",
      "value": 1.9092530678363373,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_13",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_14",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.628742552009303",
      "value": 2.628742552009303,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_14",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_15",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "1.6510177398142527",
      "value": 1.6510177398142527,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_15",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_16",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.037206093517459",
      "value": 2.037206093517459,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_16",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_17",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.845023265317937",
      "value": 1.845023265317937,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_17",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_18",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "3.7083936407207694",
      "value": 3.7083936407207694,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_18",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_19",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "4.264705882352941",
      "value": 4.264705882352941,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_19",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_20",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.5371517027863777",
      "value": 1.5371517027863777,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_20",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_21",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.3359133126934986",
      "value": 1.3359133126934986,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_21",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_22",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "13.320433436532507",
      "value": 13.320433436532507,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_22",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_23",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "3.6344905447360993",
      "value": 3.6344905447360993,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_23",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_24",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.4479254868755291",
      "value": 1.4479254868755291,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_24",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_25",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.4635901778154106",
      "value": 1.4635901778154106,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_25",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_26",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_summary",
      "notes": "; Full MDX original source independently verified; unchanged metric definition, no rescoring.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "13.697149308495625",
      "value": 13.697149308495625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_26",
        "aggregate_source_id": "aggregate-source-0164",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_new_objective_0_objective_answer_brier_sum_mean",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.8744595432347092",
      "value": 0.8744595432347092,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_new_objective_0_objective_answer_brier_sum_mean",
        "aggregate_source_id": "aggregate-source-0165",
        "source_field": "/runs/0/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_new_objective_0_actual_objective_accuracy_check",
      "task": "rlwm",
      "model": "llama8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.417455727542477",
      "value": 0.417455727542477,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_new_objective_0_actual_objective_accuracy_check",
        "aggregate_source_id": "aggregate-source-0165",
        "source_field": "/runs/0/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_new_objective_1_objective_answer_brier_sum_mean",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.2660081538304445",
      "value": 0.2660081538304445,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_new_objective_1_objective_answer_brier_sum_mean",
        "aggregate_source_id": "aggregate-source-0165",
        "source_field": "/runs/1/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_new_objective_1_actual_objective_accuracy_check",
      "task": "false_belief_story_qa",
      "model": "llama8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.8183183183183184",
      "value": 0.8183183183183184,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_new_objective_1_actual_objective_accuracy_check",
        "aggregate_source_id": "aggregate-source-0165",
        "source_field": "/runs/1/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_existing_objective_0_objective_answer_brier_sum_mean",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.709712127765174",
      "value": 0.709712127765174,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_existing_objective_0_objective_answer_brier_sum_mean",
        "aggregate_source_id": "aggregate-source-0166",
        "source_field": "/runs/0/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_existing_objective_0_actual_objective_accuracy_check",
      "task": "rlwm",
      "model": "laya",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.34850976167526604",
      "value": 0.34850976167526604,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_existing_objective_0_actual_objective_accuracy_check",
        "aggregate_source_id": "aggregate-source-0166",
        "source_field": "/runs/0/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_existing_objective_1_objective_answer_brier_sum_mean",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.2911880940594198",
      "value": 0.2911880940594198,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_existing_objective_1_objective_answer_brier_sum_mean",
        "aggregate_source_id": "aggregate-source-0166",
        "source_field": "/runs/1/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_existing_objective_1_actual_objective_accuracy_check",
      "task": "rlwm",
      "model": "intern_decision_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.7779478747330423",
      "value": 0.7779478747330423,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_existing_objective_1_actual_objective_accuracy_check",
        "aggregate_source_id": "aggregate-source-0166",
        "source_field": "/runs/1/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_existing_objective_2_objective_answer_brier_sum_mean",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.5158587110819134",
      "value": 0.5158587110819134,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_existing_objective_2_objective_answer_brier_sum_mean",
        "aggregate_source_id": "aggregate-source-0166",
        "source_field": "/runs/2/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_existing_objective_2_actual_objective_accuracy_check",
      "task": "rlwm",
      "model": "qwen35_4b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.5931847819924023",
      "value": 0.5931847819924023,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_existing_objective_2_actual_objective_accuracy_check",
        "aggregate_source_id": "aggregate-source-0166",
        "source_field": "/runs/2/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_0_invalid_rate",
      "task": "nonstationary_bandit",
      "model": "laya",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all_callbacks",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "1600",
      "n": 1600,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_0_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/0/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_1_invalid_rate",
      "task": "nonstationary_bandit",
      "model": "intern_decision_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all_callbacks",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "1600",
      "n": 1600,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_1_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/1/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_2_invalid_rate",
      "task": "nonstationary_bandit",
      "model": "llama8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all_callbacks",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "1600",
      "n": 1600,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_2_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/2/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_3_invalid_rate",
      "task": "nonstationary_bandit",
      "model": "minitaur8b",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all_callbacks",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "1600",
      "n": 1600,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_3_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/3/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_4_invalid_rate",
      "task": "nonstationary_bandit",
      "model": "qwen35_4b",
      "protocol": "documented_task_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all_callbacks",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "1600",
      "n": 1600,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_4_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/4/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_5_invalid_rate",
      "task": "hybrid_information_sampling",
      "model": "laya",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_5_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/5/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_6_invalid_rate",
      "task": "hybrid_information_sampling",
      "model": "intern_decision_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_6_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/6/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_7_invalid_rate",
      "task": "hybrid_information_sampling",
      "model": "llama8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_7_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/7/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_8_invalid_rate",
      "task": "hybrid_information_sampling",
      "model": "minitaur8b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_8_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/8/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_9_invalid_rate",
      "task": "hybrid_information_sampling",
      "model": "qwen35_4b",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_9_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/9/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_10_invalid_rate",
      "task": "mouselab_tree",
      "model": "laya",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_10_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/10/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_11_invalid_rate",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_11_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/11/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_12_invalid_rate",
      "task": "mouselab_tree",
      "model": "llama8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_12_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/12/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_13_invalid_rate",
      "task": "mouselab_tree",
      "model": "minitaur8b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_13_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/13/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_14_invalid_rate",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_14_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/14/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_15_invalid_rate",
      "task": "two_step",
      "model": "laya",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_15_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/15/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_16_invalid_rate",
      "task": "two_step",
      "model": "intern_decision_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_16_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/16/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_17_invalid_rate",
      "task": "two_step",
      "model": "llama8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_17_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/17/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_18_invalid_rate",
      "task": "two_step",
      "model": "minitaur8b",
      "protocol": "cogbench_generated_100x20",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_18_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/18/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_19_invalid_rate",
      "task": "two_step",
      "model": "qwen35_4b",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "invalid_rate",
      "metric_variant": "original_adapter_output_invalid_rate",
      "scope": "all",
      "target": "original_legal_interface",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "4000",
      "n": 4000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_19_invalid_rate",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/19/metrics/invalid_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_20_accuracy",
      "task": "instrumental_learning",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.7385416666666667",
      "value": 0.7385416666666667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_20_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/20/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_21_accuracy",
      "task": "instrumental_learning",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.8135416666666667",
      "value": 0.8135416666666667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_21_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/21/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_22_accuracy",
      "task": "instrumental_learning",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.7916666666666666",
      "value": 0.7916666666666666,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_22_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/22/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_23_accuracy",
      "task": "instrumental_learning",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.8666666666666667",
      "value": 0.8666666666666667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_23_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/23/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_24_accuracy",
      "task": "instrumental_learning",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.8",
      "value": 0.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_24_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/24/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_25_accuracy",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.5479838709677419",
      "value": 0.5479838709677419,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_25_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/25/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_25_pseudo_regret",
      "task": "horizon",
      "model": "laya",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "latent_mean_pseudo_regret_per_free_choice",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "original_reward_units_per_free_choice",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "6.328801843317972",
      "value": 6.328801843317972,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_25_pseudo_regret",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/25/metrics/pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_26_accuracy",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.6644009216589861",
      "value": 0.6644009216589861,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_26_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/26/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_26_pseudo_regret",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "latent_mean_pseudo_regret_per_free_choice",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "original_reward_units_per_free_choice",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "4.350979262672811",
      "value": 4.350979262672811,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_26_pseudo_regret",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/26/metrics/pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_27_accuracy",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.7165322580645161",
      "value": 0.7165322580645161,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_27_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/27/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_27_pseudo_regret",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "namazova_generative_31_timelines",
      "mode": "autonomous_generated_environment",
      "metric": "pseudo_regret",
      "metric_variant": "latent_mean_pseudo_regret_per_free_choice",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "original_reward_units_per_free_choice",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "3.367338709677419",
      "value": 3.367338709677419,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_27_pseudo_regret",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/27/metrics/pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_28_accuracy",
      "task": "stationary_multiarmed_bandit",
      "model": "laya",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.17592628205128205",
      "value": 0.17592628205128205,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_28_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/28/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_29_accuracy",
      "task": "stationary_multiarmed_bandit",
      "model": "llama8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.1354198717948718",
      "value": 0.1354198717948718,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_29_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/29/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_30_accuracy",
      "task": "stationary_multiarmed_bandit",
      "model": "minitaur8b",
      "protocol": "fullbench_summary_history_SH",
      "mode": "autonomous_summary_history",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "target": "maximal_original_latent_arm_mean",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.09756730769230769",
      "value": 0.09756730769230769,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_30_accuracy",
        "aggregate_source_id": "aggregate-source-0167",
        "source_field": "/results/30/metrics/accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_mcj_objective_answer_brier_sum_mean",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "original_objective_labelled_responses",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Original fulltext/scalar presentation retained; not relabeled compact. Accuracy is same-prediction crosscheck.",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.8265439826245864",
      "value": 0.8265439826245864,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_mcj_objective_answer_brier_sum_mean",
        "aggregate_source_id": "aggregate-source-0168",
        "source_field": "/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_saved_mcj_actual_objective_accuracy_check",
      "task": "multiple_cue_judgment",
      "model": "minitaur8b",
      "protocol": "psych101-extended-full-prefix-typed-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "original_objective_labelled_responses",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "Original fulltext/scalar presentation retained; not relabeled compact. Accuracy is same-prediction crosscheck.",
      "n_raw": "5336",
      "n": 5336,
      "value_raw": "0.3002248875562219",
      "value": 0.3002248875562219,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_saved_mcj_actual_objective_accuracy_check",
        "aggregate_source_id": "aggregate-source-0168",
        "source_field": "/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_saved_event_answer_brier",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "event_answer_brier",
      "metric_variant": "mean(sum_k(p_k-onehot_k)^2)",
      "scope": "grammar_and_logic_objective_answer_rows",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "; Original source scorer over all108248 saved outputs. All timings are original individual batch_size1 scalar calls; shared GPU. Brier only77320 objective answers.",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.8047706199286264",
      "value": 0.8047706199286264,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_saved_event_answer_brier",
        "aggregate_source_id": "aggregate-source-0169",
        "source_field": "/groups/all/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_saved_invalid_rate",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_response)/expected_responses",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_responses",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "; Original source scorer over all108248 saved outputs. All timings are original individual batch_size1 scalar calls; shared GPU. Brier only77320 objective answers.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_saved_invalid_rate",
        "aggregate_source_id": "aggregate-source-0169",
        "source_field": "/groups/all/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_saved_latency_p50",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p50",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_median",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Original source scorer over all108248 saved outputs. All timings are original individual batch_size1 scalar calls; shared GPU. Brier only77320 objective answers.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.14658888550047777",
      "value": 0.14658888550047777,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_saved_latency_p50",
        "aggregate_source_id": "aggregate-source-0169",
        "source_field": "/groups/all/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "new_mdx_minitaur_self_saved_latency_p95",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "latency_p95",
      "metric_variant": "complete_saved_scalar_calls",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "nearest_rank_ceil_0.95n",
      "value_status": "finite",
      "status": "complete_original_scoring",
      "notes": "CPU extraction from every saved scalar call; includes first call, no warmup removal. Prefill/sequence/readout scope follows recorded adapter. No batch-wall division.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Original source scorer over all108248 saved outputs. All timings are original individual batch_size1 scalar calls; shared GPU. Brier only77320 objective answers.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.3289980669503166",
      "value": 0.3289980669503166,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "new_mdx_minitaur_self_saved_latency_p95",
        "aggregate_source_id": "aggregate-source-0169",
        "source_field": "/groups/all/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_1",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.25468547878350717",
      "value": 0.25468547878350717,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_1",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_2",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.25474360984789085",
      "value": 0.25474360984789085,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_2",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_3",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.2553055385335052",
      "value": 0.2553055385335052,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_3",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_4",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2553708640426533",
      "value": 0.2553708640426533,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_4",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_5",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.4615976053331359",
      "value": 1.4615976053331359,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_5",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_6",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.4617539575177372",
      "value": 1.4617539575177372,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_6",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.15242601496484542",
      "value": 0.15242601496484542,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_8",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.15234436255292458",
      "value": 0.15234436255292458,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_8",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_9",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.15157588425691806",
      "value": 0.15157588425691806,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_9",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_10",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.15148188803512624",
      "value": 0.15148188803512624,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_10",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_11",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.1816453139641607",
      "value": 0.1816453139641607,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_11",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_12",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.18182530970675867",
      "value": 0.18182530970675867,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_12",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_13",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.1814647419677628",
      "value": 0.1814647419677628,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_13",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_14",
      "task": "wcst",
      "model": "laya",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.18166849615806804",
      "value": 0.18166849615806804,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_14",
        "aggregate_source_id": "aggregate-source-0170",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_15",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.27895815104761945",
      "value": 0.27895815104761945,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_15",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_16",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.27928493021797085",
      "value": 0.27928493021797085,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_16",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_17",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.28902367946577917",
      "value": 0.28902367946577917,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_17",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_18",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.28908577701113375",
      "value": 0.28908577701113375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_18",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_19",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.3780772077330319",
      "value": 1.3780772077330319,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_19",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_20",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.377492913179103",
      "value": 1.377492913179103,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_20",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_21",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.21246250735432023",
      "value": 0.21246250735432023,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_21",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_22",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2121687313783911",
      "value": 0.2121687313783911,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_22",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_23",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.22114151541206345",
      "value": 0.22114151541206345,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_23",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_24",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.22055825623333855",
      "value": 0.22055825623333855,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_24",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_25",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.158396209523212",
      "value": 0.158396209523212,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_25",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_26",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.15885212482358477",
      "value": 0.15885212482358477,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_26",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_27",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.16088757768274192",
      "value": 0.16088757768274192,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_27",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_28",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.16128273482828917",
      "value": 0.16128273482828917,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_28",
        "aggregate_source_id": "aggregate-source-0171",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_29",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.3139030190672591",
      "value": 0.3139030190672591,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_29",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_30",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.3143327583503215",
      "value": 0.3143327583503215,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_30",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_31",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.3255558582025881",
      "value": 0.3255558582025881,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_31",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_32",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.3260153677277717",
      "value": 0.3260153677277717,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_32",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_33",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.4307050334551046",
      "value": 1.4307050334551046,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_33",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_34",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.4306021561560678",
      "value": 1.4306021561560678,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_34",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_35",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.288157966810707",
      "value": 0.288157966810707,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_35",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_36",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2902618786263133",
      "value": 0.2902618786263133,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_36",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_37",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.22675895732209633",
      "value": 0.22675895732209633,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_37",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_38",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.22667398463227223",
      "value": 0.22667398463227223,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_38",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_39",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.13779026430255414",
      "value": 0.13779026430255414,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_39",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_40",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.13721185510428102",
      "value": 0.13721185510428102,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_40",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_41",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.17231576372569007",
      "value": 0.17231576372569007,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_41",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_42",
      "task": "wcst",
      "model": "llama8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.17304375098008468",
      "value": 0.17304375098008468,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_42",
        "aggregate_source_id": "aggregate-source-0172",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_43",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.28155142111511694",
      "value": 0.28155142111511694,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_43",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_44",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2816371334483299",
      "value": 0.2816371334483299,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_44",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_45",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.3006092268765138",
      "value": 0.3006092268765138,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_45",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_46",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.300768386388584",
      "value": 0.300768386388584,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_46",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_47",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.3435348184735838",
      "value": 1.3435348184735838,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_47",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_48",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.3423392670155254",
      "value": 1.3423392670155254,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_48",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_49",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.21845648433867665",
      "value": 0.21845648433867665,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_49",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_50",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.21804923945428886",
      "value": 0.21804923945428886,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_50",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_51",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.21554997964385",
      "value": 0.21554997964385,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_51",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_52",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.21554022267523915",
      "value": 0.21554022267523915,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_52",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_53",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.14605572737165445",
      "value": 0.14605572737165445,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_53",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_54",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.14685588834875332",
      "value": 0.14685588834875332,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_54",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_55",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.16671230140964605",
      "value": 0.16671230140964605,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_55",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_56",
      "task": "wcst",
      "model": "minitaur8b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.1670064293554963",
      "value": 0.1670064293554963,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_56",
        "aggregate_source_id": "aggregate-source-0173",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_57",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.25195673566592697",
      "value": 0.25195673566592697,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_57",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/generative/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_58",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.25199937274580525",
      "value": 0.25199937274580525,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_58",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/generative/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_59",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.2507455430375036",
      "value": 0.2507455430375036,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_59",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/accuracy/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_60",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "accuracy",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "original_rule_correct_card",
      "unit": "fraction",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.2508232711306257",
      "value": 0.2508232711306257,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_60",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/accuracy/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_61",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.9677260157375163",
      "value": 1.9677260157375163,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_61",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/human_nll_legal/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_62",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.966466703761432",
      "value": 1.966466703761432,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_62",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/human_nll_legal/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_63",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.15879183012722714",
      "value": 0.15879183012722714,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_63",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/generative/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_64",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.15846009095185823",
      "value": 0.15846009095185823,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_64",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/generative/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_65",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.15765513492661962",
      "value": 0.15765513492661962,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_65",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/perseveration_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_66",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "perseveration_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "perseveration_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.1575976164340599",
      "value": 0.1575976164340599,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_66",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/perseveration_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_67",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.1788058880874803",
      "value": 0.1788058880874803,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_67",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/generative/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_68",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "generative",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.17915947937901835",
      "value": 0.17915947937901835,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_68",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/generative/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_69",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "participant_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.17831790328372707",
      "value": 0.17831790328372707,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_69",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/set_loss_error/participant_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_70",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "set_loss_error",
      "metric_variant": "trial_weighted_mean",
      "scope": "predictive",
      "target": "set_loss_error",
      "unit": "dimensionless",
      "aggregation": "trial_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.17853222518425593",
      "value": 0.17853222518425593,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_70",
        "aggregate_source_id": "aggregate-source-0174",
        "source_field": "/metrics/predictive/set_loss_error/trial_weighted_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_71",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4902194250722912",
      "value": 0.4902194250722912,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_71",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_72",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4902194250722912",
      "value": 0.4902194250722912,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_72",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/objective_label_accuracy_under_human_history"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_73",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4739468163519873",
      "value": 0.4739468163519873,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_73",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/human_response_agreement"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_74",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_agreement",
      "metric_variant": "human_response_agreement",
      "scope": "all",
      "target": "observed_human_response_not_objective_accuracy",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4739468163519873",
      "value": 0.4739468163519873,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_74",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/human_response_agreement"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_75",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "same_original_human_history_response_no_official_heldout_split",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "1.079768295316208",
      "value": 1.079768295316208,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_75",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/typed_choice_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_76",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "target": "same_original_human_history_response_no_official_heldout_split",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "1.079768295316208",
      "value": 1.079768295316208,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_76",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/typed_choice_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_77",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "CollinsFrank_2012_EJoN:learning/three_button_choice",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_77",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/by_experiment_phase_response_kind/CollinsFrank_2012_EJoN:learning~1three_button_choice/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_78",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_78",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_79",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.0487537209992297",
      "value": 0.0487537209992297,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_79",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_80",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "all original scalar calls on shared device; includes first call; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.09157489400240593",
      "value": 0.09157489400240593,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_80",
        "aggregate_source_id": "aggregate-source-0175",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_81",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "ToMi/test",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.6501501501501501",
      "value": 0.6501501501501501,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_81",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/by_experiment/ToMi~1test/original_answer_exact_match_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_82",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.6501501501501501",
      "value": 0.6501501501501501,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_82",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/original_answer_exact_match_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_83",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "ToMi/test",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_83",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/by_experiment/ToMi~1test/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_84",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_84",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_85",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.03042309450393077",
      "value": 0.03042309450393077,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_85",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/by_experiment/ToMi~1test/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_86",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.03042309450393077",
      "value": 0.03042309450393077,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_86",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_87",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "ToMi/test",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.03083419280119415",
      "value": 0.03083419280119415,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_87",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/by_experiment/ToMi~1test/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_88",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.03083419280119415",
      "value": 0.03083419280119415,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_88",
        "aggregate_source_id": "aggregate-source-0176",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_89",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "1.0083505201908776",
      "value": 1.0083505201908776,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_89",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_90",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_90",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_91",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Yoon2020Polite/exp2",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_91",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_92",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_92",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_93",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.03272343549906509",
      "value": 0.03272343549906509,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_93",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_94",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.03568000500308699",
      "value": 0.03568000500308699,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_94",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_95",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.03558263150262064",
      "value": 0.03558263150262064,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_95",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_96",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "32",
      "n": 32,
      "value_raw": "0.180503305849197",
      "value": 0.180503305849197,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_96",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/by_experiment/Yoon2020Polite~1exp1/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_97",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Yoon2020Polite/exp2",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "96",
      "n": 96,
      "value_raw": "0.04414051149615261",
      "value": 0.04414051149615261,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_97",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/by_experiment/Yoon2020Polite~1exp2/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_98",
      "task": "politeness_pragmatics",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "0.044147758696271924",
      "value": 0.044147758696271924,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_98",
        "aggregate_source_id": "aggregate-source-0177",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_99",
      "task": "pragmatic_language_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "1.0486177484416257",
      "value": 1.0486177484416257,
      "common_metric": false,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_99",
        "aggregate_source_id": "aggregate-source-0178",
        "source_field": "/human_response_weighted_mean_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_100",
      "task": "pragmatic_language_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Hu2023Fine/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_100",
        "aggregate_source_id": "aggregate-source-0178",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_101",
      "task": "pragmatic_language_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_101",
        "aggregate_source_id": "aggregate-source-0178",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_102",
      "task": "pragmatic_language_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.03573096850232105",
      "value": 0.03573096850232105,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_102",
        "aggregate_source_id": "aggregate-source-0178",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_103",
      "task": "pragmatic_language_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.03573096850232105",
      "value": 0.03573096850232105,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_103",
        "aggregate_source_id": "aggregate-source-0178",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_104",
      "task": "pragmatic_language_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Hu2023Fine/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.04778246119858522",
      "value": 0.04778246119858522,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_104",
        "aggregate_source_id": "aggregate-source-0178",
        "source_field": "/by_experiment/Hu2023Fine~1exp1/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_105",
      "task": "pragmatic_language_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "0.04778246119858522",
      "value": 0.04778246119858522,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_105",
        "aggregate_source_id": "aggregate-source-0178",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_106",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Re-read on 2026-10-31 from the saved completions. The earlier parser rejected text after the closing marker and reported 1.0. No model call.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_106",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/invalid_response_rate",
        "reread_from": "results/aggregate/minitaur_nonliteral_reparsed.csv"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_107",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Re-read on 2026-10-31 from the saved completions. The earlier parser rejected text after the closing marker and reported 1.0. No model call.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_107",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/invalid_response_rate",
        "reread_from": "results/aggregate/minitaur_nonliteral_reparsed.csv"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_108",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Re-read on 2026-10-31 from the saved completions. The earlier parser rejected text after the closing marker and reported 1.0. No model call.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_108",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/invalid_response_rate",
        "reread_from": "results/aggregate/minitaur_nonliteral_reparsed.csv"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_109",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.4018490180023946",
      "value": 0.4018490180023946,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_109",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_110",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.34962307649766444",
      "value": 0.34962307649766444,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_110",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_111",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.401002698999946",
      "value": 0.401002698999946,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_111",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_112",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp1",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "300",
      "n": 300,
      "value_raw": "0.40889774605493584",
      "value": 0.40889774605493584,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_112",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp1/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_113",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "Tsvilodub2025Nonliteral/exp3b",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.3678665042039938",
      "value": 0.3678665042039938,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_113",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/by_experiment/Tsvilodub2025Nonliteral~1exp3b/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_114",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "recorded_request_quantile",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Shared-device wall latency, not human reaction time.; Shared host/device; not exclusive latency or human reaction time; includes prefill/readout according to source.; Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "0.4088537671035738",
      "value": 0.4088537671035738,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_114",
        "aggregate_source_id": "aggregate-source-0179",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_115",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "4.049976210409848",
      "value": 4.049976210409848,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_115",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_116",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "3.814130653567946",
      "value": 3.814130653567946,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_116",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_117",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "3.814130653567946",
      "value": 3.814130653567946,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_117",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_118",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "4.107150890856381",
      "value": 4.107150890856381,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_118",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_119",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "4.107150890856381",
      "value": 4.107150890856381,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_119",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_120",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "32.97560975609756",
      "value": 32.97560975609756,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_120",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/all/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_121",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "35.625",
      "value": 35.625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_121",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_122",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "35.625",
      "value": 35.625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_122",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_123",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "32.333333333333336",
      "value": 32.333333333333336,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_123",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_124",
      "task": "probability_coherence_judgment",
      "model": "laya",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "32.333333333333336",
      "value": 32.333333333333336,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_124",
        "aggregate_source_id": "aggregate-source-0180",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_125",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "3.1714039925476887",
      "value": 3.1714039925476887,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_125",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_126",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "2.687496942733557",
      "value": 2.687496942733557,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_126",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_127",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "2.687496942733557",
      "value": 2.687496942733557,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_127",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_128",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.28871479250263",
      "value": 3.28871479250263,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_128",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_129",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.28871479250263",
      "value": 3.28871479250263,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_129",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_130",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "18.945528455284553",
      "value": 18.945528455284553,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_130",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/all/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_131",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "23.8375",
      "value": 23.8375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_131",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_132",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "23.8375",
      "value": 23.8375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_132",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_133",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "17.75959595959596",
      "value": 17.75959595959596,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_133",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_134",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "17.75959595959596",
      "value": 17.75959595959596,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_134",
        "aggregate_source_id": "aggregate-source-0181",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_135",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "3.7328601890042696",
      "value": 3.7328601890042696,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_135",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_136",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "3.2612748862595793",
      "value": 3.2612748862595793,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_136",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_137",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "3.2612748862595793",
      "value": 3.2612748862595793,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_137",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_138",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.847183898760557",
      "value": 3.847183898760557,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_138",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_139",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.847183898760557",
      "value": 3.847183898760557,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_139",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_140",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "28.567073170731707",
      "value": 28.567073170731707,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_140",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/all/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_141",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "29.8875",
      "value": 29.8875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_141",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_142",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "29.8875",
      "value": 29.8875,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_142",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_143",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "28.246969696969696",
      "value": 28.246969696969696,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_143",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_144",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "28.246969696969696",
      "value": 28.246969696969696,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_144",
        "aggregate_source_id": "aggregate-source-0182",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_145",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "3.0695631442864832",
      "value": 3.0695631442864832,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_145",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_146",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "2.699260530206185",
      "value": 2.699260530206185,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_146",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_147",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "2.699260530206185",
      "value": 2.699260530206185,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_147",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_148",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.159333474972613",
      "value": 3.159333474972613,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_148",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_149",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.159333474972613",
      "value": 3.159333474972613,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_149",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_150",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "19.04471544715447",
      "value": 19.04471544715447,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_150",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/all/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_151",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "23.895833333333332",
      "value": 23.895833333333332,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_151",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_152",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "23.895833333333332",
      "value": 23.895833333333332,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_152",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_153",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "17.86868686868687",
      "value": 17.86868686868687,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_153",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_154",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "17.86868686868687",
      "value": 17.86868686868687,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_154",
        "aggregate_source_id": "aggregate-source-0183",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_155",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "3.290387847549567",
      "value": 3.290387847549567,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_155",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_156",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "2.7532562844148214",
      "value": 2.7532562844148214,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_156",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_157",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "2.7532562844148214",
      "value": 2.7532562844148214,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_157",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_158",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.4206015598246498",
      "value": 3.4206015598246498,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_158",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_159",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "3.4206015598246498",
      "value": 3.4206015598246498,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_159",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_160",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "21.86869918699187",
      "value": 21.86869918699187,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_160",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/all/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_161",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "25.35625",
      "value": 25.35625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_161",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_162",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp1.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "480",
      "n": 480,
      "value_raw": "25.35625",
      "value": 25.35625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_162",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp1.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_163",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "21.023232323232325",
      "value": 21.023232323232325,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_163",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_164",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "zhu2020bayesian/exp2.csv:integer_percentage",
      "target": "original_human_numeric_response",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "1980",
      "n": 1980,
      "value_raw": "21.023232323232325",
      "value": 21.023232323232325,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_164",
        "aggregate_source_id": "aggregate-source-0184",
        "source_field": "/groups/zhu2020bayesian~1exp2.csv:integer_percentage/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_165",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.4561562338334196",
      "value": 0.4561562338334196,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_165",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/all/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_166",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.5148606811145511",
      "value": 0.5148606811145511,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_166",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_167",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.5148606811145511",
      "value": 0.5148606811145511,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_167",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_168",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.4508044030482642",
      "value": 0.4508044030482642,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_168",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_169",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.4508044030482642",
      "value": 0.4508044030482642,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_169",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_170",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "2.028509628820405",
      "value": 2.028509628820405,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_170",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_171",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "9044",
      "n": 9044,
      "value_raw": "1.953456241788366",
      "value": 1.953456241788366,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_171",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_172",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.658323334345726",
      "value": 2.658323334345726,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_172",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_173",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "1.6047066214557923",
      "value": 1.6047066214557923,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_173",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_174",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.9712176364635456",
      "value": 1.9712176364635456,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_174",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_175",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.1635079417363",
      "value": 2.1635079417363,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_175",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_176",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "4.508272257933415",
      "value": 4.508272257933415,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_176",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_177",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "99204",
      "n": 99204,
      "value_raw": "2.0353519217956295",
      "value": 2.0353519217956295,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_177",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_178",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.782572199482356",
      "value": 2.782572199482356,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_178",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_179",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "1.715959450929454",
      "value": 1.715959450929454,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_179",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_180",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.016286942577649",
      "value": 2.016286942577649,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_180",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_181",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.0684763153529557",
      "value": 2.0684763153529557,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_181",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_182",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "4.467996938429618",
      "value": 4.467996938429618,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_182",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_183",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.91640866873065",
      "value": 2.91640866873065,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_183",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_184",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.3219814241486068",
      "value": 1.3219814241486068,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_184",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_185",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.9427244582043344",
      "value": 1.9427244582043344,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_185",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_186",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "33.62538699690403",
      "value": 33.62538699690403,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_186",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_187",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "4.005503810330229",
      "value": 4.005503810330229,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_187",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_188",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.4357888794806661",
      "value": 1.4357888794806661,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_188",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_189",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.7919841941857184",
      "value": 1.7919841941857184,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_189",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_190",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "39.19390347163421",
      "value": 39.19390347163421,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_190",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_191",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.2649120538023797",
      "value": 0.2649120538023797,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_191",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/all/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_192",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.3848297213622291",
      "value": 0.3848297213622291,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_192",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_193",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.3848297213622291",
      "value": 0.3848297213622291,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_193",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_194",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.2539796782387807",
      "value": 0.2539796782387807,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_194",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_195",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.2539796782387807",
      "value": 0.2539796782387807,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_195",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_196",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "2.523089865473279",
      "value": 2.523089865473279,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_196",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_197",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "9044",
      "n": 9044,
      "value_raw": "2.236208361312558",
      "value": 2.236208361312558,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_197",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_198",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "3.3489304701796727",
      "value": 3.3489304701796727,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_198",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_199",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "1.777029363968341",
      "value": 1.777029363968341,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_199",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_200",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.2077686596834707",
      "value": 2.2077686596834707,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_200",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_201",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.7785654147931993",
      "value": 2.7785654147931993,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_201",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_202",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "5.201358874035905",
      "value": 5.201358874035905,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_202",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_203",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "99204",
      "n": 99204,
      "value_raw": "2.5492436125361877",
      "value": 2.5492436125361877,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_203",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_204",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "3.553814570738377",
      "value": 3.553814570738377,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_204",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_205",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "2.2045718426010974",
      "value": 2.2045718426010974,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_205",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_206",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.46927272018941",
      "value": 2.46927272018941,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_206",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_207",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.870520369894337",
      "value": 2.870520369894337,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_207",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_208",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "4.750084488671724",
      "value": 4.750084488671724,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_208",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_209",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "6.7105263157894735",
      "value": 6.7105263157894735,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_209",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_210",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.4969040247678018",
      "value": 1.4969040247678018,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_210",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_211",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.6656346749226005",
      "value": 1.6656346749226005,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_211",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_212",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "24.270897832817337",
      "value": 24.270897832817337,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_212",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_213",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "6.090601185436071",
      "value": 6.090601185436071,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_213",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_214",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.7393451876940447",
      "value": 1.7393451876940447,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_214",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_215",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.9404459497600903",
      "value": 1.9404459497600903,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_215",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_216",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "19.37510584250635",
      "value": 19.37510584250635,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_216",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_217",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.189187790998448",
      "value": 0.189187790998448,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_217",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/all/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_218",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.7001547987616099",
      "value": 0.7001547987616099,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_218",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_219",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.7001547987616099",
      "value": 0.7001547987616099,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_219",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_220",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.14260513688964155",
      "value": 0.14260513688964155,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_220",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_221",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.14260513688964155",
      "value": 0.14260513688964155,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_221",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_222",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "1.8851206979439843",
      "value": 1.8851206979439843,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_222",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_223",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "9044",
      "n": 9044,
      "value_raw": "1.6204117613224445",
      "value": 1.6204117613224445,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_223",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_224",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.594243082954053",
      "value": 2.594243082954053,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_224",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_225",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "1.2642765432512604",
      "value": 1.2642765432512604,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_225",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_226",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.951911625847153",
      "value": 1.951911625847153,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_226",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_227",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.7579043065700388",
      "value": 1.7579043065700388,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_227",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_228",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "3.738940210630389",
      "value": 3.738940210630389,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_228",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_229",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "99204",
      "n": 99204,
      "value_raw": "1.9092530678363373",
      "value": 1.9092530678363373,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_229",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_230",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.628742552009303",
      "value": 2.628742552009303,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_230",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_231",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "1.6510177398142527",
      "value": 1.6510177398142527,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_231",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_232",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.037206093517459",
      "value": 2.037206093517459,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_232",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_233",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.845023265317937",
      "value": 1.845023265317937,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_233",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_234",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "3.7083936407207694",
      "value": 3.7083936407207694,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_234",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_235",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "4.264705882352941",
      "value": 4.264705882352941,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_235",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_236",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.5371517027863777",
      "value": 1.5371517027863777,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_236",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_237",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.3359133126934986",
      "value": 1.3359133126934986,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_237",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_238",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "13.320433436532507",
      "value": 13.320433436532507,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_238",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_239",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "3.6344905447360993",
      "value": 3.6344905447360993,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_239",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_240",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.4479254868755291",
      "value": 1.4479254868755291,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_240",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_241",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.4635901778154106",
      "value": 1.4635901778154106,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_241",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_242",
      "task": "performance_self_assessment",
      "model": "minitaur8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "13.697149308495625",
      "value": 13.697149308495625,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_242",
        "aggregate_source_id": "aggregate-source-0187",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_243",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "all",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "77320",
      "n": 77320,
      "value_raw": "0.5570356958096223",
      "value": 0.5570356958096223,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_243",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/all/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_244",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.42213622291021674",
      "value": 0.42213622291021674,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_244",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_245",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.42213622291021674",
      "value": 0.42213622291021674,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_245",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_246",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.5693338978267005",
      "value": 0.5693338978267005,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_246",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_247",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_accuracy",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "original_objective_answer_or_optimal_action_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.5693338978267005",
      "value": 0.5693338978267005,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_247",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/objective_accuracy"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_248",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "2.261557447575329",
      "value": 2.261557447575329,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_248",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/all/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_249",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "9044",
      "n": 9044,
      "value_raw": "2.2494971655607854",
      "value": 2.2494971655607854,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_249",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_250",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.7110852839918813",
      "value": 2.7110852839918813,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_250",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_251",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "1.933657009615895",
      "value": 1.933657009615895,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_251",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_252",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.287149382115192",
      "value": 2.287149382115192,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_252",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_253",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "2.505378166525375",
      "value": 2.505378166525375,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_253",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_254",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "4.652777389059533",
      "value": 4.652777389059533,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_254",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_255",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "99204",
      "n": 99204,
      "value_raw": "2.262656931371769",
      "value": 2.262656931371769,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_255",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_256",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "3.073926008782281",
      "value": 3.073926008782281,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_256",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_257",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:answer_choice",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "1.929868839954638",
      "value": 1.929868839954638,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_257",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:answer_choice/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_258",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.1590348794007377",
      "value": 2.1590348794007377,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_258",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_259",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "2.1697077734112162",
      "value": 2.1697077734112162,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_259",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_260",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "observed_human_response",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "4.975839978067262",
      "value": 4.975839978067262,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_260",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_typed_nll"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_261",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "3.0897832817337463",
      "value": 3.0897832817337463,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_261",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_262",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "1.3854489164086687",
      "value": 1.3854489164086687,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_262",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_263",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "3.289473684210526",
      "value": 3.289473684210526,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_263",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_264",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021dunningkruger/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "646",
      "n": 646,
      "value_raw": "32.139318885448915",
      "value": 32.139318885448915,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_264",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021dunningkruger~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_265",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:absAssess",
      "target": "original_human_numeric_response",
      "unit": "number_correct_0_to_20",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "4.065057860570138",
      "value": 4.065057860570138,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_265",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:absAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_266",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffOther",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.466130397967824",
      "value": 1.466130397967824,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_266",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffOther/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_267",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:diffSelf",
      "target": "original_human_numeric_response",
      "unit": "difficulty_rating_0_to_10",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "1.7856336438046854",
      "value": 1.7856336438046854,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_267",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:diffSelf/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_268",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "jansen2021logic/exp1.csv:relAssess",
      "target": "original_human_numeric_response",
      "unit": "percentile_rank_0_to_100",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete original sample count; actual model-specific readout retained. Shared-device timing is not exclusive speed.",
      "n_raw": "7086",
      "n": 7086,
      "value_raw": "34.5386677956534",
      "value": 34.5386677956534,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_268",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/jansen2021logic~1exp1.csv:relAssess/mean_numeric_absolute_error"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_281",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_response)/expected_responses",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_responses",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete merged source summary invalid-response rate; original row denominator retained.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_281",
        "aggregate_source_id": "aggregate-source-0185",
        "source_field": "/groups/all/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_286",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_response)/expected_responses",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_responses",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete merged source summary invalid-response rate; original row denominator retained.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_286",
        "aggregate_source_id": "aggregate-source-0186",
        "source_field": "/groups/all/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7613",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "invalid_rate",
      "metric_variant": "sum(invalid_response)/expected_responses",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "fraction",
      "aggregation": "all_required_responses",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Complete merged source summary invalid-response rate; original row denominator retained.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7613",
        "aggregate_source_id": "aggregate-source-0188",
        "source_field": "/groups/all/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7614",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.; Original class-summed objective-answer Brier/selected-answer accuracy on all original labelled rows, no class-count normalization.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.6153964305748164",
      "value": 0.6153964305748164,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7614",
        "aggregate_source_id": "aggregate-source-0189",
        "source_field": "/runs/0/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7615",
      "task": "rlwm",
      "model": "minitaur8b",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Unchanged original objective-answer scorer; accuracy is a crosscheck of the same predictions, not an independent replicate. No class-count division.; Original class-summed objective-answer Brier/selected-answer accuracy on all original labelled rows, no class-count normalization.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.4902194250722912",
      "value": 0.4902194250722912,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7615",
        "aggregate_source_id": "aggregate-source-0189",
        "source_field": "/runs/0/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7616",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "event_answer_brier",
      "metric_variant": "objective_answer_brier_sum_mean",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "sum_over_classes_Brier",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Original class-summed objective-answer Brier/selected-answer accuracy on all original labelled rows, no class-count normalization.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.4442157975996006",
      "value": 0.4442157975996006,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7616",
        "aggregate_source_id": "aggregate-source-0189",
        "source_field": "/runs/1/objective_answer_brier_sum_mean"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7617",
      "task": "false_belief_story_qa",
      "model": "minitaur8b",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "actual_objective_accuracy_check",
      "scope": "objective_labelled_including_recovered_prior_learning_labels",
      "target": "original_objective_answer_label",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Original class-summed objective-answer Brier/selected-answer accuracy on all original labelled rows, no class-count normalization.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.6501501501501501",
      "value": 0.6501501501501501,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7617",
        "aggregate_source_id": "aggregate-source-0189",
        "source_field": "/runs/1/actual_objective_accuracy_check"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7618",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_reward_free_additional_metric",
      "scope": "native_autonomous_full",
      "target": "original_free_choice_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "free_choice_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "54.9275633640553",
      "value": 54.9275633640553,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7618",
        "aggregate_source_id": "aggregate-source-0190",
        "source_field": "/mean_reward_free_additional_metric"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7619",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_reward_free_additional_metric",
      "scope": "native_autonomous_full",
      "target": "original_free_choice_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "free_choice_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "52.22911866359447",
      "value": 52.22911866359447,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7619",
        "aggregate_source_id": "aggregate-source-0191",
        "source_field": "/mean_reward_free_additional_metric"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7620",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "native_autonomous_full",
      "target": "whole_original_decision_latency",
      "unit": "seconds",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.037956892500915274",
      "value": 0.037956892500915274,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7620",
        "aggregate_source_id": "aggregate-source-0192",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7621",
      "task": "horizon",
      "model": "llama8b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "native_autonomous_full",
      "target": "whole_original_decision_latency",
      "unit": "seconds",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.03850833559936291",
      "value": 0.03850833559936291,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7621",
        "aggregate_source_id": "aggregate-source-0192",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7622",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "native_autonomous_full",
      "target": "best_path_expected_value_at_STOP",
      "unit": "task_specific_reward_units",
      "aggregation": "episode_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7622",
        "aggregate_source_id": "aggregate-source-0193",
        "source_field": "/mean_terminal_expected_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7623",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "native_autonomous_full",
      "target": "whole_original_decision_latency",
      "unit": "seconds",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.05229703300028632",
      "value": 0.05229703300028632,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7623",
        "aggregate_source_id": "aggregate-source-0193",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7624",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "native_autonomous_full",
      "target": "whole_original_decision_latency",
      "unit": "seconds",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0.05318603630021244",
      "value": 0.05318603630021244,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7624",
        "aggregate_source_id": "aggregate-source-0193",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7625",
      "task": "mouselab_tree",
      "model": "intern_decision_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "native_autonomous_full",
      "target": "invalid_typed_response",
      "unit": "proportion",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7625",
        "aggregate_source_id": "aggregate-source-0193",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7626",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "native_autonomous_full",
      "target": "best_path_expected_value_at_STOP",
      "unit": "task_specific_reward_units",
      "aggregation": "episode_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "17.40063157894737",
      "value": 17.40063157894737,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7626",
        "aggregate_source_id": "aggregate-source-0194",
        "source_field": "/mean_terminal_expected_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7627",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50_seconds",
      "scope": "native_autonomous_full",
      "target": "whole_original_decision_latency",
      "unit": "seconds",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "392905",
      "n": 392905,
      "value_raw": "0.05139841200070805",
      "value": 0.05139841200070805,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7627",
        "aggregate_source_id": "aggregate-source-0194",
        "source_field": "/latency_p50_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7628",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95_seconds",
      "scope": "native_autonomous_full",
      "target": "whole_original_decision_latency",
      "unit": "seconds",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "392905",
      "n": 392905,
      "value_raw": "0.05225534999772208",
      "value": 0.05225534999772208,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7628",
        "aggregate_source_id": "aggregate-source-0194",
        "source_field": "/latency_p95_seconds"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7629",
      "task": "mouselab_tree",
      "model": "qwen35_4b",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "invalid_rate",
      "metric_variant": "invalid_response_rate",
      "scope": "native_autonomous_full",
      "target": "invalid_typed_response",
      "unit": "proportion",
      "aggregation": "all_decision_calls",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Cost=0, expected path value; not realized route payoff.; Strict all-original-ID/environment/action/feedback/scoring receipt. Expected STOP value, not realized route payoff; original zero cost and action rule retained.",
      "n_raw": "392905",
      "n": 392905,
      "value_raw": "0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7629",
        "aggregate_source_id": "aggregate-source-0194",
        "source_field": "/invalid_response_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7630",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "accuracy",
      "metric_variant": "optimal_action_rate",
      "scope": "free_choices",
      "target": "latent_mean_optimal_arm",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "HOU29471 all74400 original rows accepted; strict greater latent-mean optimal-choice definition retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.786520737327189",
      "value": 0.786520737327189,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7630",
        "aggregate_source_id": "aggregate-source-0195",
        "source_field": "/horizon_independent_scalar_acceptance/neural_metrics_updated/0/optimal_action_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7631",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "reward_per_free_choice",
      "scope": "free_choices",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "HOU29471 all74400 original rows accepted; strict greater latent-mean optimal-choice definition retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "54.9275633640553",
      "value": 54.9275633640553,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7631",
        "aggregate_source_id": "aggregate-source-0195",
        "source_field": "/horizon_independent_scalar_acceptance/neural_metrics_updated/0/reward_per_free_choice"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7632",
      "task": "horizon",
      "model": "intern_decision_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_per_free_choice",
      "scope": "free_choices",
      "target": "maximal_original_latent_arm_mean",
      "unit": "original_reward_units_per_free_choice",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.; HOU29471 all74400 original rows accepted; strict greater latent-mean optimal-choice definition retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "2.2247119815668204",
      "value": 2.2247119815668204,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7632",
        "aggregate_source_id": "aggregate-source-0195",
        "source_field": "/horizon_independent_scalar_acceptance/neural_metrics_updated/0/pseudo_regret_per_free_choice"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7633",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "accuracy",
      "metric_variant": "optimal_action_rate",
      "scope": "free_choices",
      "target": "latent_mean_optimal_arm",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "HOU29471 all74400 original rows accepted; strict greater latent-mean optimal-choice definition retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.6325460829493088",
      "value": 0.6325460829493088,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7633",
        "aggregate_source_id": "aggregate-source-0195",
        "source_field": "/horizon_independent_scalar_acceptance/neural_metrics_updated/1/optimal_action_rate"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7634",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "reward_per_free_choice",
      "scope": "free_choices",
      "target": "autonomous_environment_reward",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "HOU29471 all74400 original rows accepted; strict greater latent-mean optimal-choice definition retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "52.22911866359447",
      "value": 52.22911866359447,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7634",
        "aggregate_source_id": "aggregate-source-0195",
        "source_field": "/horizon_independent_scalar_acceptance/neural_metrics_updated/1/reward_per_free_choice"
      },
      "runtime_context": {}
    },
    {
      "row_id": "final20261028_7635",
      "task": "horizon",
      "model": "qwen35_4b",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret_per_free_choice",
      "scope": "free_choices",
      "target": "maximal_original_latent_arm_mean",
      "unit": "original_reward_units_per_free_choice",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_source_readback",
      "notes": "Original specified metric computed from complete saved outputs. Named source protocol and mode retained; no model call or new experiment.; HOU29471 all74400 original rows accepted; strict greater latent-mean optimal-choice definition retained.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "4.872119815668203",
      "value": 4.872119815668203,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "final20261028_7635",
        "aggregate_source_id": "aggregate-source-0195",
        "source_field": "/horizon_independent_scalar_acceptance/neural_metrics_updated/1/pseudo_regret_per_free_choice"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_1",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_complete_integer_sequence",
      "mode": "human_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "10.129616981983418",
      "value": 10.129616981983418,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_1",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/0/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_2",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_complete_integer_sequence",
      "mode": "human_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "12.01202416518936",
      "value": 12.01202416518936,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_2",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/1/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_3",
      "task": "probability_coherence_judgment",
      "model": "intern_decision_4b",
      "protocol": "original_complete_integer_sequence",
      "mode": "human_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_3",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/2/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_4",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "1.1679795050004032",
      "value": 1.1679795050004032,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_4",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/3/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_5",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "2.2226977280515707",
      "value": 2.2226977280515707,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_5",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/4/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_6",
      "task": "probability_coherence_judgment",
      "model": "llama8b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_6",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/5/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_7",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "1.4847334290097933",
      "value": 1.4847334290097933,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_7",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/6/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_8",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "2.8421619325381466",
      "value": 2.8421619325381466,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_8",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/7/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_9",
      "task": "probability_coherence_judgment",
      "model": "minitaur8b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_9",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/8/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_10",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "8.362584306014469",
      "value": 8.362584306014469,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_10",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/9/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_11",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "10.07709432033298",
      "value": 10.07709432033298,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_11",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/10/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_12",
      "task": "probability_coherence_judgment",
      "model": "qwen35_4b",
      "protocol": "original_scalar_numeric_readout",
      "mode": "human_prediction",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original per-response batch_size=1 throughout; shared-device wall time, not human reaction time or isolated model speed.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_12",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/11/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_13",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "all",
      "target": "original_ground_rule_correct_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Four-class sum, not divided by class count; predictive and autonomous histories remain distinct.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.7753336176191803",
      "value": 0.7753336176191803,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_13",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/12/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_14",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_14",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/13/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_15",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.662224663014058",
      "value": 0.662224663014058,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_15",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/14/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_16",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.1733566982526098",
      "value": 1.1733566982526098,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_16",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/15/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_17",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "all",
      "target": "original_ground_rule_correct_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Four-class sum, not divided by class count; predictive and autonomous histories remain distinct.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.9002309829860197",
      "value": 0.9002309829860197,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_17",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/16/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_18",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_18",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/17/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_19",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.6375489545171149",
      "value": 0.6375489545171149,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_19",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/18/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_20",
      "task": "wcst",
      "model": "intern_decision_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.1490082024305592",
      "value": 1.1490082024305592,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_20",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/19/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_21",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "all",
      "target": "original_ground_rule_correct_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Four-class sum, not divided by class count; predictive and autonomous histories remain distinct.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.0171519924242827",
      "value": 1.0171519924242827,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_21",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/20/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_22",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_22",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/21/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_23",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.600086298014503",
      "value": 0.600086298014503,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_23",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/22/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_24",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.1263026287400861",
      "value": 1.1263026287400861,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_24",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/23/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_25",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "all",
      "target": "original_ground_rule_correct_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Four-class sum, not divided by class count; predictive and autonomous histories remain distinct.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "1.2495973476897198",
      "value": 1.2495973476897198,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_25",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/24/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_26",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_26",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/25/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_27",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.4618654674850404",
      "value": 0.4618654674850404,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_27",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/26/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_28",
      "task": "wcst",
      "model": "qwen35_4b",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Original single-trial timing on a shared device; no hardware speed ranking.",
      "n_raw": "12754",
      "n": 12754,
      "value_raw": "0.8845865877985487",
      "value": 0.8845865877985487,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_28",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/27/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_29",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full_original_history_objective_answer_probabilities",
      "mode": "human_history_objective_scoring",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Class-summed answer probability score; grammar and logic retained separately. Numeric self-reports are not event probabilities. Classes=[5]",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.5959014564884616",
      "value": 0.5959014564884616,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_29",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/28/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_30",
      "task": "performance_self_assessment",
      "model": "intern_decision_4b",
      "protocol": "full_original_history_objective_answer_probabilities",
      "mode": "human_history_objective_scoring",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Class-summed answer probability score; grammar and logic retained separately. Numeric self-reports are not event probabilities. Classes=[5]",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.6809237980446821",
      "value": 0.6809237980446821,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_30",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/29/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_31",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full_original_history_objective_answer_probabilities",
      "mode": "human_history_objective_scoring",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Class-summed answer probability score; grammar and logic retained separately. Numeric self-reports are not event probabilities. Classes=[5]",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.7664046103902383",
      "value": 0.7664046103902383,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_31",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/30/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_32",
      "task": "performance_self_assessment",
      "model": "llama8b",
      "protocol": "full_original_history_objective_answer_probabilities",
      "mode": "human_history_objective_scoring",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Class-summed answer probability score; grammar and logic retained separately. Numeric self-reports are not event probabilities. Classes=[5]",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.9511210227988216",
      "value": 0.9511210227988216,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_32",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/31/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_33",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full_original_history_objective_answer_probabilities",
      "mode": "human_history_objective_scoring",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "jansen2021dunningkruger/exp1.csv",
      "target": "original_objective_answer_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Class-summed answer probability score; grammar and logic retained separately. Numeric self-reports are not event probabilities. Classes=[5]",
      "n_raw": "6460",
      "n": 6460,
      "value_raw": "0.7556197895341015",
      "value": 0.7556197895341015,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_33",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/32/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_34",
      "task": "performance_self_assessment",
      "model": "qwen35_4b",
      "protocol": "full_original_history_objective_answer_probabilities",
      "mode": "human_history_objective_scoring",
      "metric": "event_answer_brier",
      "metric_variant": "event_answer_brier",
      "scope": "jansen2021logic/exp1.csv",
      "target": "original_objective_answer_key",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "Class-summed answer probability score; grammar and logic retained separately. Numeric self-reports are not event probabilities. Classes=[5]",
      "n_raw": "70860",
      "n": 70860,
      "value_raw": "0.578812464316147",
      "value": 0.578812464316147,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_34",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/33/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_35",
      "task": "nonliteral_number_interpretation",
      "model": "minitaur8b",
      "protocol": "original_public_330_items_numeric_readout",
      "mode": "original_source_scenarios",
      "metric": "mae",
      "metric_variant": "mae",
      "scope": "all",
      "target": "original_human_native_scale_report",
      "unit": "original_response_units",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete",
      "notes": "All 330 responses were re-read on 2026-10-31 from the saved completions, taking the number before the closing marker. Generation had run past the marker and the earlier parser rejected the trailing text, so they were wrongly counted as invalid. No model call.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "28.52972727272727",
      "value": 28.52972727272727,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_35",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/34/value",
        "reread_from": "results/aggregate/minitaur_nonliteral_reparsed.csv"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_36",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "original_native_closed_loop_timing_34720",
      "mode": "generative",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Full original31 timelines and34720 free choices; shared-device timing.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.03571483499990791",
      "value": 0.03571483499990791,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_36",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/35/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_37",
      "task": "horizon",
      "model": "minitaur8b",
      "protocol": "original_native_closed_loop_timing_34720",
      "mode": "generative",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "complete",
      "notes": "Full original31 timelines and34720 free choices; shared-device timing.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.03784847254996748",
      "value": 0.03784847254996748,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_37",
        "aggregate_source_id": "aggregate-source-0196",
        "source_field": "/metrics/36/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_38",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "223.8",
      "value": 223.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_38",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/0/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_39",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_39",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/1/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_40",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "231.2059237899257",
      "value": 231.2059237899257,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_40",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/2/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_41",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "37.068026246997",
      "value": 37.068026246997,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_41",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/3/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_42",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "399.1",
      "value": 399.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_42",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/4/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_43",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999994",
      "value": 199.99999999999994,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_43",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/5/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_44",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "5371.657638285617",
      "value": 5371.657638285617,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_44",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/6/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_45",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "21.489920863833742",
      "value": 21.489920863833742,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_45",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/7/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_46",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "213.1",
      "value": 213.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_46",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/8/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_47",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "13.483333333333333",
      "value": 13.483333333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_47",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/9/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_48",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "224.96042521817873",
      "value": 224.96042521817873,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_48",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/10/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_49",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "383.06127333287657",
      "value": 383.06127333287657,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_49",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/11/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_50",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "554.3333333333334",
      "value": 554.3333333333334,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_50",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/12/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_51",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "47.99333333333332",
      "value": 47.99333333333332,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_51",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/13/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_52",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4883.790113985455",
      "value": 4883.790113985455,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_52",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/14/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_53",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "113.63392122406836",
      "value": 113.63392122406836,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_53",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/15/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_54",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "72.5",
      "value": 72.5,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_54",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/16/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_55",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_55",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/17/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_56",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "477.06742269357",
      "value": 477.06742269357,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_56",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/18/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_57",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "10.332495219815124",
      "value": 10.332495219815124,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_57",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/19/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_58",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "415.1333333333333",
      "value": 415.1333333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_58",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/20/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_59",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "186.97999999999993",
      "value": 186.97999999999993,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_59",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/21/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_60",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4853.944656347249",
      "value": 4853.944656347249,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_60",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/22/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_61",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "548.2161891484367",
      "value": 548.2161891484367,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_61",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/23/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_62",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "73.8",
      "value": 73.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_62",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/24/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_63",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_63",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/25/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_64",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "62.035393679406404",
      "value": 62.035393679406404,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_64",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/26/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_65",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "337.61581362862364",
      "value": 337.61581362862364,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_65",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/27/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_66",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "400.8",
      "value": 400.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_66",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/28/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_67",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999994",
      "value": 199.99999999999994,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_67",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/29/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_68",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "2644.9097380352387",
      "value": 2644.9097380352387,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_68",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/30/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_69",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "1885.826350477169",
      "value": 1885.826350477169,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_69",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/31/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_70",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "accuracy",
      "metric_variant": "accuracy",
      "scope": "all_16_original_conditions",
      "target": "maximal_original_latent_arm_mean",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.3736025641025641",
      "value": 0.3736025641025641,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_70",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/32/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_71",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all_16_original_conditions",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "all_required_steps",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_71",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/33/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_72",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.07584051799494773",
      "value": 0.07584051799494773,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_72",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/34/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_73",
      "task": "stationary_multiarmed_bandit",
      "model": "intern_decision_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.10580542330972091",
      "value": 0.10580542330972091,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_73",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/35/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_74",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "223.16666666666666",
      "value": 223.16666666666666,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_74",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/36/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_75",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "00_VideoWatching_b_arms5_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "0.5166666666666667",
      "value": 0.5166666666666667,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_75",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/37/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_76",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "213.48442512291868",
      "value": 213.48442512291868,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_76",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/38/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_77",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "54.789524914004005",
      "value": 54.789524914004005,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_77",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/39/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_78",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "400.8333333333333",
      "value": 400.8333333333333,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_78",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/40/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_79",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "02_VideoWatching_b_arms5_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "198.25999999999993",
      "value": 198.25999999999993,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_79",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/41/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_80",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4668.514763933077",
      "value": 4668.514763933077,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_80",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/42/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_81",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "724.6327952163763",
      "value": 724.6327952163763,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_81",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/43/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_82",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "204.0",
      "value": 204.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_82",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/44/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_83",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "04_ClothesShopping_b_arms5_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "22.283333333333335",
      "value": 22.283333333333335,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_83",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/45/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_84",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "484.1985626008136",
      "value": 484.1985626008136,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_84",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/46/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_85",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "123.82313595024185",
      "value": 123.82313595024185,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_85",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/47/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_86",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "401.1",
      "value": 401.1,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_86",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/48/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_87",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "06_ClothesShopping_b_arms5_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999994",
      "value": 199.99999999999994,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_87",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/49/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_88",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "4896.308892770226",
      "value": 4896.308892770226,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_88",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/50/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_89",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "101.11514243929766",
      "value": 101.11514243929766,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_89",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/51/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_90",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "72.5",
      "value": 72.5,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_90",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/52/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_91",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "08_VideoWatching_b_arms20_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_91",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/53/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_92",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "295.1362996025392",
      "value": 295.1362996025392,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_92",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/54/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_93",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "192.26361831084594",
      "value": 192.26361831084594,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_93",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/55/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_94",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "402.8",
      "value": 402.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_94",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/56/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_95",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "10_VideoWatching_b_arms20_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999994",
      "value": 199.99999999999994,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_95",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/57/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_96",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "1403.9932987112927",
      "value": 1403.9932987112927,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_96",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/58/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_97",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "3998.1675467843947",
      "value": 3998.1675467843947,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_97",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/59/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_98",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "73.8",
      "value": 73.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_98",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/60/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_99",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "12_ClothesShopping_b_arms20_difficulty_0_5",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "150.0",
      "value": 150.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_99",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/61/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_100",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "107.29166568599346",
      "value": 107.29166568599346,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_100",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/62/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_101",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "292.3595416220365",
      "value": 292.3595416220365,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_101",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/63/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_102",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "400.8",
      "value": 400.8,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_102",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/64/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_103",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "14_ClothesShopping_b_arms20_difficulty_0_2",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "199.99999999999994",
      "value": 199.99999999999994,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_103",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/65/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_104",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "reward",
      "metric_variant": "reward",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "original_cumulative_reward",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "2576.0711486285622",
      "value": 2576.0711486285622,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_104",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/66/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_105",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "pseudo_regret",
      "metric_variant": "pseudo_regret",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "original_latent_mean_pseudo_regret",
      "unit": "original_environment_reward_units",
      "aggregation": "mean_episode_cumulative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Original16 conditions and30 trajectories per condition; reward scales remain separate. Public-code adaptation, not paper-matched horizons/decoding.",
      "n_raw": "30",
      "n": 30,
      "value_raw": "1954.664939883845",
      "value": 1954.664939883845,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_105",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/67/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_106",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "accuracy",
      "metric_variant": "accuracy",
      "scope": "all_16_original_conditions",
      "target": "maximal_original_latent_arm_mean",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "value_status": "finite",
      "status": "complete_saved_output_scoring",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.2890641025641026",
      "value": 0.2890641025641026,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_106",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/68/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_107",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "invalid_rate",
      "metric_variant": "invalid_rate",
      "scope": "all_16_original_conditions",
      "target": "invalid_responses_over_all_required_responses",
      "unit": "proportion",
      "aggregation": "all_required_steps",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_107",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/69/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_108",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "latency_p50",
      "metric_variant": "latency_p50",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.08705151898902841",
      "value": 0.08705151898902841,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_108",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/70/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "cpuclose20261028_109",
      "task": "stationary_multiarmed_bandit",
      "model": "qwen35_4b",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "latency_p95",
      "metric_variant": "latency_p95",
      "scope": "all_16_original_conditions",
      "target": "model_request_wall_time",
      "unit": "seconds",
      "aggregation": "sample_quantile",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Full saved312000 decision rows. Original saved per-request timings exclude CPU replay. Shared-device/source-machine timings do not establish isolated model speed; full manifest retained.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.15458490286546292",
      "value": 0.15458490286546292,
      "common_metric": true,
      "source": {
        "asset": "final_numeric_long_24_closeout_20261028.csv",
        "original_row_id": "cpuclose20261028_109",
        "aggregate_source_id": "aggregate-source-0197",
        "source_field": "/metrics/71/value"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-080-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 232.657418961422,
      "value_raw": "232.657418961422",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 80,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-080-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "01_VideoWatching_g_arms5_difficulty_0.00",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 35.616531075500674,
      "value_raw": "35.616531075500674",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 80,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-081-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 5275.510629817991,
      "value_raw": "5275.510629817991",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 81,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-081-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "03_VideoWatching_g_arms5_difficulty_1.30",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 117.63692933146075,
      "value_raw": "117.63692933146075",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 81,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-082-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 571.3655613658863,
      "value_raw": "571.3655613658863",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 82,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-082-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "05_ClothesShopping_g_arms5_difficulty_1.32",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 36.656137185169115,
      "value_raw": "36.656137185169115",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 82,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-083-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 4892.891262834581,
      "value_raw": "4892.891262834581",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 83,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-083-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "07_ClothesShopping_g_arms5_difficulty_0.50",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 104.53277237494298,
      "value_raw": "104.53277237494298",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 83,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-084-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 347.02385155540645,
      "value_raw": "347.02385155540645",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 84,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-084-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "09_VideoWatching_g_arms20_difficulty_0.00",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 140.37606635797874,
      "value_raw": "140.37606635797874",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 84,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-085-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 4967.724584610339,
      "value_raw": "4967.724584610339",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 85,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-085-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "11_VideoWatching_g_arms20_difficulty_0.00",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 434.43626088534734,
      "value_raw": "434.43626088534734",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 85,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-086-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 244.62797211624033,
      "value_raw": "244.62797211624033",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 86,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-086-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "13_ClothesShopping_g_arms20_difficulty_0.21",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 9000,
      "value": 155.0232351917896,
      "value_raw": "155.0232351917896",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 86,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-087-reward",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "reward",
      "metric_variant": "mean_cumulative_reward",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "sum_of_original_environment_observed_rewards",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 4006.2768263007692,
      "value_raw": "4006.2768263007692",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 87,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_reward"
      },
      "runtime_context": {}
    },
    {
      "row_id": "gaussian-ts-condition-087-pseudo_regret",
      "task": "stationary_multiarmed_bandit",
      "model": "gaussian_ts",
      "model_kind": "disclosed_classical_reimplementation",
      "protocol": "fullbench_original_gaussian_NIG_standard_disclosed_reimplementation",
      "mode": "autonomous_classical_policy",
      "metric": "pseudo_regret",
      "metric_variant": "mean_cumulative_pseudo_regret",
      "scope": "15_ClothesShopping_g_arms20_difficulty_0.00",
      "target": "sum_of_latent_mean_optimal_reward_minus_chosen_reward",
      "unit": "condition_specific_reward_units",
      "aggregation": "mean_episode_cumulative_total",
      "n": 30,
      "n_raw": "30",
      "steps": 30000,
      "value": 524.4592622116372,
      "value_raw": "524.4592622116372",
      "value_status": "finite",
      "status": "complete_full_original_environment_and_independent_posterior_acceptance",
      "common_metric": true,
      "notes": "Eight Gaussian conditions only; 30 original trajectory seeds per condition. Disclosed Normal-Inverse-Gamma classical policy; not claimed to be the missing author implementation. Reward and pseudo-regret are not pooled across reward families, arm counts or horizons.",
      "source": {
        "asset": "full_condition_scores.csv",
        "original_row_index": 87,
        "aggregate_source_id": "accepted-gaussian-ts-summary",
        "source_field": "mean_cumulative_pseudo_regret"
      },
      "runtime_context": {}
    },
    {
      "row_id": "a00001",
      "task": "bayesian_urn_inference",
      "model": "jev_1_14",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "target": "analytic_Bayes_posterior",
      "mode_key": "bayesian_urn_inference",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.0682182908004872",
      "value": 0.0682182908004872,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/Jev4_full_results_20261030.json",
        "field": "/rows/0/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 400
      },
      "table2_marker": null
    },
    {
      "row_id": "a00002",
      "task": "bayesian_urn_inference",
      "model": "claude_opus_5_5",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "target": "analytic_Bayes_posterior",
      "mode_key": "bayesian_urn_inference",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Opus 5.5). Readout: generated FINAL line.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.000970842717169546",
      "value": 0.000970842717169546,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/BAYESIAN400_ACCEPTED_RESULTS_20261029.json",
        "field": "/rows/0/posterior_mae_all400"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 400
      },
      "table2_marker": null
    },
    {
      "row_id": "a00003",
      "task": "bayesian_urn_inference",
      "model": "claude_sonnet_5_5",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "target": "analytic_Bayes_posterior",
      "mode_key": "bayesian_urn_inference",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Readout: generated FINAL line.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.000970842717169546",
      "value": 0.000970842717169546,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/BAYESIAN400_ACCEPTED_RESULTS_20261029.json",
        "field": "/rows/1/posterior_mae_all400"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 400
      },
      "table2_marker": null
    },
    {
      "row_id": "a00004",
      "task": "bayesian_urn_inference",
      "model": "claude_haiku_5_5",
      "protocol": "cogbench_generated_posterior_200x2",
      "mode": "generated_explicit_posterior",
      "metric": "mae",
      "metric_variant": "posterior_mae",
      "scope": "all",
      "unit": "probability_0_to_1",
      "aggregation": "response_weighted_mean",
      "target": "analytic_Bayes_posterior",
      "mode_key": "bayesian_urn_inference",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.000970842717169546",
      "value": 0.000970842717169546,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/BAYESIAN400_ACCEPTED_RESULTS_20261029.json",
        "field": "/rows/2/posterior_mae_all400"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 400
      },
      "table2_marker": null
    },
    {
      "row_id": "a00005",
      "task": "multiple_cue_judgment",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multiple_cue_judgment",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/multiple_cue_judgment/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 6136
      },
      "table2_marker": null
    },
    {
      "row_id": "a00006",
      "task": "multiple_cue_judgment",
      "model": "claude_opus_5_5",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multiple_cue_judgment",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00007",
      "task": "multiple_cue_judgment",
      "model": "claude_sonnet_5_5",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multiple_cue_judgment",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00008",
      "task": "multiple_cue_judgment",
      "model": "claude_haiku_5_5",
      "protocol": "compact_full_history",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multiple_cue_judgment",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.a. 63 saved decisions; full sample 6136. calibration-only rows of the four-row verbalized-probability quote ; no full run.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/claude_verbalized_probability_20261030/calib4/records/multiple_cue_judgment/responses.jsonl",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00009",
      "task": "multi_attribute_choice",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multi_attribute_choice",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.3291398970428249",
      "value": 0.3291398970428249,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/multi_attribute_choice/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 576
      },
      "table2_marker": null
    },
    {
      "row_id": "a00010",
      "task": "multi_attribute_choice",
      "model": "claude_opus_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multi_attribute_choice",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00011",
      "task": "multi_attribute_choice",
      "model": "claude_sonnet_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multi_attribute_choice",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00012",
      "task": "multi_attribute_choice",
      "model": "claude_haiku_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multi_attribute_choice",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: stated probability.",
      "n_raw": "576",
      "n": 576,
      "value_raw": "0.2055455720779773",
      "value": 0.2055455720779773,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/claude_verbalized_probability_20261030/RESULTS.json",
        "field": "/tasks/multi_attribute_choice/strict_nll"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "stated probability",
        "model_calls": 576
      },
      "table2_marker": null
    },
    {
      "row_id": "a00013",
      "task": "probability_coherence_judgment",
      "model": "jev_1_14",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "original_human_numeric_response",
      "mode_key": "probability_coherence_judgment",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "2460",
      "n": 2460,
      "value_raw": "17.65569105691057",
      "value": 17.65569105691057,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/table2_followup/JEV_TWO_ADDITIONS_ACCEPTED_20261030.json",
        "field": "/models_or_tasks/coherence/summary/primary_mae_original_percentage_points"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 2460
      },
      "table2_marker": null
    },
    {
      "row_id": "a00014",
      "task": "probability_coherence_judgment",
      "model": "claude_opus_5_5",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "original_human_numeric_response",
      "mode_key": "probability_coherence_judgment",
      "value_status": "undefined",
      "status": "cancelled_before_generation",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints canc. 0 saved decisions; full sample 2460. canceled before generation.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/api_results_20261029/final_api_table2/FINAL_API_RESULTS_ONLINE_20261030.json",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "canc."
    },
    {
      "row_id": "a00015",
      "task": "probability_coherence_judgment",
      "model": "claude_sonnet_5_5",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "original_human_numeric_response",
      "mode_key": "probability_coherence_judgment",
      "value_status": "undefined",
      "status": "cancelled_before_generation",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints canc. 0 saved decisions; full sample 2460. canceled before generation.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/api_results_20261029/final_api_table2/FINAL_API_RESULTS_ONLINE_20261030.json",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "canc."
    },
    {
      "row_id": "a00016",
      "task": "probability_coherence_judgment",
      "model": "claude_haiku_5_5",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "mae",
      "metric_variant": "mean_numeric_absolute_error",
      "scope": "all",
      "unit": "probability_judgment_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "original_human_numeric_response",
      "mode_key": "probability_coherence_judgment",
      "value_status": "undefined",
      "status": "partial_records_retained",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints part. 181 saved decisions; full sample 2460. partial prefix: stopped by the authors on 2026-10-30.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/claude_coherence_20261030/budget100_20261030/CLOSED_RECORD_SOURCE_ACCEPTANCE.json",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "part."
    },
    {
      "row_id": "a00017",
      "task": "shj_category_learning",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "shj_category_learning",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.8741812227074236",
      "value": 0.8741812227074236,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/shj_category_learning/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 3664
      },
      "table2_marker": null
    },
    {
      "row_id": "a00018",
      "task": "shj_category_learning",
      "model": "claude_opus_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "shj_category_learning",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00019",
      "task": "shj_category_learning",
      "model": "claude_sonnet_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "shj_category_learning",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00020",
      "task": "shj_category_learning",
      "model": "claude_haiku_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "category_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "shj_category_learning",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "3664",
      "n": 3664,
      "value_raw": "0.9123908296943232",
      "value": 0.9123908296943232,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/claude-haiku-5-5/MODEL_ACCEPTANCE.json",
        "field": "/tasks/shj_category_learning/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 3664
      },
      "table2_marker": null
    },
    {
      "row_id": "a00021",
      "task": "medin_category_learning",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "medin_category_learning",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "6300",
      "n": 6300,
      "value_raw": "0.8680952380952381",
      "value": 0.8680952380952381,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/medin_category_learning/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 6300
      },
      "table2_marker": null
    },
    {
      "row_id": "a00022",
      "task": "medin_category_learning",
      "model": "claude_opus_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "medin_category_learning",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00023",
      "task": "medin_category_learning",
      "model": "claude_sonnet_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "medin_category_learning",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00024",
      "task": "medin_category_learning",
      "model": "claude_haiku_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "medin_category_learning",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "6300",
      "n": 6300,
      "value_raw": "0.9741269841269842",
      "value": 0.9741269841269842,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/claude-haiku-5-5/MODEL_ACCEPTANCE.json",
        "field": "/tasks/medin_category_learning/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 6300
      },
      "table2_marker": null
    },
    {
      "row_id": "a00025",
      "task": "weather_prediction",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "weather_prediction",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.8075",
      "value": 0.8075,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/weather_prediction/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 400
      },
      "table2_marker": null
    },
    {
      "row_id": "a00026",
      "task": "weather_prediction",
      "model": "claude_opus_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "weather_prediction",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00027",
      "task": "weather_prediction",
      "model": "claude_sonnet_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "weather_prediction",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00028",
      "task": "weather_prediction",
      "model": "claude_haiku_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "weather_prediction",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "400",
      "n": 400,
      "value_raw": "0.8025",
      "value": 0.8025,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/claude-haiku-5-5/MODEL_ACCEPTANCE.json",
        "field": "/tasks/weather_prediction/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 400
      },
      "table2_marker": null
    },
    {
      "row_id": "a00029",
      "task": "rlwm",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "rlwm",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.8946154863827938",
      "value": 0.8946154863827938,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_big_tasks_20261030/environment_tasks/jev-1.14/rlwm/ACCEPTANCE.json",
        "field": "/table2_value_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 52911
      },
      "table2_marker": null
    },
    {
      "row_id": "a00030",
      "task": "rlwm",
      "model": "claude_opus_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "rlwm",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00031",
      "task": "rlwm",
      "model": "claude_sonnet_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "rlwm",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00032",
      "task": "rlwm",
      "model": "claude_haiku_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "original_human_history_replay_no_official_heldout_split",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "rlwm",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "52911",
      "n": 52911,
      "value_raw": "0.8939728978851279",
      "value": 0.8939728978851279,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_big_tasks_20261030/haiku_calibration/claude-haiku-5-5/rlwm/ACCEPTANCE.json",
        "field": "/table2_value_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 52911
      },
      "table2_marker": null
    },
    {
      "row_id": "a00033",
      "task": "recent_probes",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "recent_probes",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "0.9904610492845787",
      "value": 0.9904610492845787,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/recent_probes/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 3774
      },
      "table2_marker": null
    },
    {
      "row_id": "a00034",
      "task": "recent_probes",
      "model": "claude_opus_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "recent_probes",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00035",
      "task": "recent_probes",
      "model": "claude_sonnet_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "recent_probes",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00036",
      "task": "recent_probes",
      "model": "claude_haiku_5_5",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "accuracy",
      "metric_variant": "objective_label_accuracy_under_human_history",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "recent_probes",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "3774",
      "n": 3774,
      "value_raw": "1.0",
      "value": 1.0,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/claude-haiku-5-5/MODEL_ACCEPTANCE.json",
        "field": "/tasks/recent_probes/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 3774
      },
      "table2_marker": null
    },
    {
      "row_id": "a00037",
      "task": "wcst",
      "model": "jev_1_14",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "target": "observed_human_response",
      "mode_key": "wcst:human_predictive",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "1.5433271104448631",
      "value": 1.5433271104448631,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/Jev4_full_results_20261030.json",
        "field": "/rows/4/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 12754
      },
      "table2_marker": null
    },
    {
      "row_id": "a00038",
      "task": "wcst",
      "model": "claude_opus_5_5",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "target": "observed_human_response",
      "mode_key": "wcst:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00039",
      "task": "wcst",
      "model": "claude_sonnet_5_5",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "target": "observed_human_response",
      "mode_key": "wcst:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00040",
      "task": "wcst",
      "model": "claude_haiku_5_5",
      "protocol": "wcst_predictive_75_test_participants",
      "mode": "predictive",
      "metric": "human_nll",
      "metric_variant": "participant_mean",
      "scope": "predictive",
      "unit": "nats_per_response",
      "aggregation": "participant_mean",
      "target": "observed_human_response",
      "mode_key": "wcst:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.a. 129 saved decisions; full sample 12754. calibration-only rows of the four-row verbalized-probability quote ; no full run.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/claude_verbalized_probability_20261030/calib4/records/wcst_prediction/responses.jsonl",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00041",
      "task": "wcst",
      "model": "jev_1_14",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "target": "original_rule_correct_card",
      "mode_key": "wcst:autonomous_generative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "75",
      "n": 75,
      "value_raw": "0.28498473191859414",
      "value": 0.28498473191859414,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/Jev4_full_results_20261030.json",
        "field": "/rows/5/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 12754
      },
      "table2_marker": null
    },
    {
      "row_id": "a00042",
      "task": "wcst",
      "model": "claude_opus_5_5",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "target": "original_rule_correct_card",
      "mode_key": "wcst:autonomous_generative",
      "value_status": "undefined",
      "status": "non_full_prefix_only",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.f. 200 saved decisions; full sample 12754. 200-call pilot prefix only (n.f. in Table 3).",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/probes_20261028/p2/results_claude_api_20261029/opus55_pilot_200x2",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.f."
    },
    {
      "row_id": "a00043",
      "task": "wcst",
      "model": "claude_sonnet_5_5",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "target": "original_rule_correct_card",
      "mode_key": "wcst:autonomous_generative",
      "value_status": "undefined",
      "status": "non_full_prefix_only",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.f. 200 saved decisions; full sample 12754. 200-call pilot prefix only (n.f. in Table 3).",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/probes_20261028/p2/results_claude_extension_20261029/sonnet55_pilot_200x2",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.f."
    },
    {
      "row_id": "a00044",
      "task": "wcst",
      "model": "claude_haiku_5_5",
      "protocol": "wcst_generative_75_test_participants",
      "mode": "generative",
      "metric": "accuracy",
      "metric_variant": "participant_mean",
      "scope": "generative",
      "unit": "fraction",
      "aggregation": "participant_mean",
      "target": "original_rule_correct_card",
      "mode_key": "wcst:autonomous_generative",
      "value_status": "undefined",
      "status": "non_full_prefix_only",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.f. 200 saved decisions; full sample 12754. 200-call pilot prefix only (n.f. in Table 3).",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/probes_20261028/p2/results_claude_extension_20261029/haiku55_pilot_200x2",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.f."
    },
    {
      "row_id": "a00045",
      "task": "nonstationary_bandit",
      "model": "jev_1_14",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "nonstationary_bandit",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.63875",
      "value": 0.63875,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/Jev4_full_results_20261030.json",
        "field": "/rows/1/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 1600
      },
      "table2_marker": null
    },
    {
      "row_id": "a00046",
      "task": "nonstationary_bandit",
      "model": "claude_opus_5_5",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "nonstationary_bandit",
      "value_status": "undefined",
      "status": "partial_records_retained",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints part. 436 saved decisions; full sample 1600. partial prefix: stopped by the authors on 2026-10-30.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/claude_restless_20261029/CLAUDE_RESTLESS_CLOSED_RECORD_SOURCE_ACCEPTANCE_20261029.json",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "part."
    },
    {
      "row_id": "a00047",
      "task": "nonstationary_bandit",
      "model": "claude_sonnet_5_5",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "nonstationary_bandit",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Readout: generated FINAL line.",
      "n_raw": "800",
      "n": 800,
      "value_raw": "0.91125",
      "value": 0.91125,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/final_api_table2/FINAL_API_RESULTS_ONLINE_20261030.json",
        "field": "/sonnet_restless/original_post_step_accuracy"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 1600
      },
      "table2_marker": null
    },
    {
      "row_id": "a00048",
      "task": "nonstationary_bandit",
      "model": "claude_haiku_5_5",
      "protocol": "cogbench_restless_action_and_confidence_10x80",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "accuracy_post_step_original",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "mean_of_10_equal_length_80_trial_run_scores",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "nonstationary_bandit",
      "value_status": "undefined",
      "status": "partial_records_retained",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints part. 1508 saved decisions; full sample 1600. partial: 9 of 10 runs complete; run 0 ended at an invalid response after 68 responses; no full-sample score.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/claude_restless_20261029/CLAUDE_RESTLESS_CLOSED_RECORD_SOURCE_ACCEPTANCE_20261029.json",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "part."
    },
    {
      "row_id": "a00049",
      "task": "instrumental_learning",
      "model": "jev_1_14",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "instrumental_learning",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "960",
      "n": 960,
      "value_raw": "0.8666666666666667",
      "value": 0.8666666666666667,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/environment_tasks/jev-1.14/instrumental_learning/ACCEPTANCE.json",
        "field": "/score_fraction"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 960
      },
      "table2_marker": null
    },
    {
      "row_id": "a00050",
      "task": "instrumental_learning",
      "model": "claude_opus_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "instrumental_learning",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00051",
      "task": "instrumental_learning",
      "model": "claude_sonnet_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "instrumental_learning",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00052",
      "task": "instrumental_learning",
      "model": "claude_haiku_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "accuracy",
      "metric_variant": "original_latent_mean_optimal_action_rate",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "instrumental_learning",
      "value_status": "undefined",
      "status": "partial_records_retained",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints part. 843 saved decisions; full sample 10 runs (960 decisions for Jev). partial: 8 of 10 runs complete (full_protocol_reached false); no Table 3 score.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/api_small_tasks_20261030/environment_tasks/claude-haiku-5-5/instrumental_learning/ACCEPTANCE.json",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "part."
    },
    {
      "row_id": "a00053",
      "task": "horizon",
      "model": "jev_1_14",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "horizon:human_predictive",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.41409505415082937",
      "value": 0.41409505415082937,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/Jev4_full_results_20261030.json",
        "field": "/rows/2/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 6720
      },
      "table2_marker": null
    },
    {
      "row_id": "a00054",
      "task": "horizon",
      "model": "claude_opus_5_5",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "horizon:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00055",
      "task": "horizon",
      "model": "claude_sonnet_5_5",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "horizon:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00056",
      "task": "horizon",
      "model": "claude_haiku_5_5",
      "protocol": "namazova_human_predictive_6x1120",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "legal_nll_eps1e8",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "horizon:human_predictive",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: stated probability.",
      "n_raw": "6720",
      "n": 6720,
      "value_raw": "0.38981806508734024",
      "value": 0.38981806508734024,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/claude_verbalized_probability_20261030/horizon_full/RESULTS_HORIZON.json",
        "field": "/table_value_source_epsilon_1e8/participant_mean_same_data"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "stated probability",
        "model_calls": 6720
      },
      "table2_marker": null
    },
    {
      "row_id": "a00057",
      "task": "horizon",
      "model": "jev_1_14",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "accuracy",
      "metric_variant": "optimal_action_rate",
      "scope": "free_choices",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "latent_mean_optimal_arm",
      "mode_key": "horizon:autonomous_generative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "34720",
      "n": 34720,
      "value_raw": "0.8290610599078341",
      "value": 0.8290610599078341,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/Jev4_full_results_20261030.json",
        "field": "/rows/3/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 34720
      },
      "table2_marker": null
    },
    {
      "row_id": "a00058",
      "task": "horizon",
      "model": "claude_opus_5_5",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "accuracy",
      "metric_variant": "optimal_action_rate",
      "scope": "free_choices",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "latent_mean_optimal_arm",
      "mode_key": "horizon:autonomous_generative",
      "value_status": "undefined",
      "status": "non_full_prefix_only",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.f. 200 saved decisions; full sample 34720. 200-call pilot prefix only (n.f. in Table 3).",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/probes_20261028/p2/results_claude_api_20261029/opus55_pilot_200x2",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.f."
    },
    {
      "row_id": "a00059",
      "task": "horizon",
      "model": "claude_sonnet_5_5",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "accuracy",
      "metric_variant": "optimal_action_rate",
      "scope": "free_choices",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "latent_mean_optimal_arm",
      "mode_key": "horizon:autonomous_generative",
      "value_status": "undefined",
      "status": "non_full_prefix_only",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.f. 200 saved decisions; full sample 34720. 200-call pilot prefix only (n.f. in Table 3).",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/probes_20261028/p2/results_claude_extension_20261029/sonnet55_pilot_200x2",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.f."
    },
    {
      "row_id": "a00060",
      "task": "horizon",
      "model": "claude_haiku_5_5",
      "protocol": "namazova_generative_31_timelines_native",
      "mode": "autonomous_generative",
      "metric": "accuracy",
      "metric_variant": "optimal_action_rate",
      "scope": "free_choices",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "latent_mean_optimal_arm",
      "mode_key": "horizon:autonomous_generative",
      "value_status": "undefined",
      "status": "non_full_prefix_only",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.f. 200 saved decisions; full sample 34720. 200-call pilot prefix only (n.f. in Table 3).",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/probes_20261028/p2/results_claude_extension_20261029/haiku55_pilot_200x2",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.f."
    },
    {
      "row_id": "a00061",
      "task": "stationary_multiarmed_bandit",
      "model": "jev_1_14",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "accuracy",
      "metric_variant": "accuracy",
      "scope": "all_16_original_conditions",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "stationary_multiarmed_bandit",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "312000",
      "n": 312000,
      "value_raw": "0.47033974358974356",
      "value": 0.47033974358974356,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_big_tasks_20261030/environment_tasks/jev-1.14/stationary_bandit/ACCEPTANCE.json",
        "field": "/table2_value_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 312000
      },
      "table2_marker": null
    },
    {
      "row_id": "a00062",
      "task": "stationary_multiarmed_bandit",
      "model": "claude_opus_5_5",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "accuracy",
      "metric_variant": "accuracy",
      "scope": "all_16_original_conditions",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "stationary_multiarmed_bandit",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00063",
      "task": "stationary_multiarmed_bandit",
      "model": "claude_sonnet_5_5",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "accuracy",
      "metric_variant": "accuracy",
      "scope": "all_16_original_conditions",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "stationary_multiarmed_bandit",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00064",
      "task": "stationary_multiarmed_bandit",
      "model": "claude_haiku_5_5",
      "protocol": "official_code_fullbench_summary_history_16_conditions",
      "mode": "autonomous",
      "metric": "accuracy",
      "metric_variant": "accuracy",
      "scope": "all_16_original_conditions",
      "unit": "proportion",
      "aggregation": "response_weighted_mean",
      "target": "maximal_original_latent_arm_mean",
      "mode_key": "stationary_multiarmed_bandit",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.e. 0 saved decisions; full sample 312000. dropped by the authors before any request.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/api_big_tasks_20261030/haiku_calibration/AUTHOR_APPROVED_PLAN.json",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00065",
      "task": "hybrid_information_sampling",
      "model": "jev_1_14",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "hybrid_information_sampling",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "163135",
      "n": 163135,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_big_tasks_20261030/offline_prediction/jev-1.14/hybrid_information_sampling/ACCEPTANCE.json",
        "field": "/strict_nll_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 163135
      },
      "table2_marker": null
    },
    {
      "row_id": "a00066",
      "task": "hybrid_information_sampling",
      "model": "claude_opus_5_5",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "hybrid_information_sampling",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00067",
      "task": "hybrid_information_sampling",
      "model": "claude_sonnet_5_5",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "hybrid_information_sampling",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00068",
      "task": "hybrid_information_sampling",
      "model": "claude_haiku_5_5",
      "protocol": "original_20_subjects_8fold_frozen_prediction",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "pooled_action_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "hybrid_information_sampling",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00069",
      "task": "mouselab_tree",
      "model": "jev_1_14",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "mouselab_tree:human_predictive",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "20363",
      "n": 20363,
      "value_raw": "1.4575866455484785",
      "value": 1.4575866455484785,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/mouselab_prediction/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 20363
      },
      "table2_marker": null
    },
    {
      "row_id": "a00070",
      "task": "mouselab_tree",
      "model": "claude_opus_5_5",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "mouselab_tree:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00071",
      "task": "mouselab_tree",
      "model": "claude_sonnet_5_5",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "mouselab_tree:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00072",
      "task": "mouselab_tree",
      "model": "claude_haiku_5_5",
      "protocol": "original_human_predictive_95x25",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "mouselab_tree:human_predictive",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.a. 205 saved decisions; full sample 20363. calibration-only rows of the four-row verbalized-probability quote ; no full run.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/claude_verbalized_probability_20261030/calib4/records/mouselab_prediction/responses.jsonl",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00073",
      "task": "mouselab_tree",
      "model": "jev_1_14",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "native_autonomous_full",
      "unit": "task_specific_reward_units",
      "aggregation": "episode_mean",
      "target": "best_path_expected_value_at_STOP",
      "mode_key": "mouselab_tree:autonomous_generative",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "23750",
      "n": 23750,
      "value_raw": "8.213684210526315",
      "value": 8.213684210526315,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_big_tasks_20261030/environment_tasks/jev-1.14/mouselab_autonomous/ACCEPTANCE.json",
        "field": "/table2_value_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 101739
      },
      "table2_marker": null
    },
    {
      "row_id": "a00074",
      "task": "mouselab_tree",
      "model": "claude_opus_5_5",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "native_autonomous_full",
      "unit": "task_specific_reward_units",
      "aggregation": "episode_mean",
      "target": "best_path_expected_value_at_STOP",
      "mode_key": "mouselab_tree:autonomous_generative",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00075",
      "task": "mouselab_tree",
      "model": "claude_sonnet_5_5",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "native_autonomous_full",
      "unit": "task_specific_reward_units",
      "aggregation": "episode_mean",
      "target": "best_path_expected_value_at_STOP",
      "mode_key": "mouselab_tree:autonomous_generative",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00076",
      "task": "mouselab_tree",
      "model": "claude_haiku_5_5",
      "protocol": "original_metamdp_generative_23750_native",
      "mode": "autonomous_generative",
      "metric": "reward",
      "metric_variant": "mean_terminal_expected_reward",
      "scope": "native_autonomous_full",
      "unit": "task_specific_reward_units",
      "aggregation": "episode_mean",
      "target": "best_path_expected_value_at_STOP",
      "mode_key": "mouselab_tree:autonomous_generative",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.e. 1566 saved decisions; full sample 23,750 episodes. calibration only (runs 0-237, 1,566 saved responses); the production run was not started.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/runs/api_big_tasks_20261030/haiku_calibration/claude-haiku-5-5/mouselab_autonomous",
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00077",
      "task": "bart",
      "model": "jev_1_14",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "bart",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "900",
      "n": 900,
      "value_raw": "1.761111111111111",
      "value": 1.761111111111111,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/environment_tasks/jev-1.14/bart/ACCEPTANCE.json",
        "field": "/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 2566
      },
      "table2_marker": null
    },
    {
      "row_id": "a00078",
      "task": "bart",
      "model": "claude_opus_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "bart",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00079",
      "task": "bart",
      "model": "claude_sonnet_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "bart",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00080",
      "task": "bart",
      "model": "claude_haiku_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "bart",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "900",
      "n": 900,
      "value_raw": "6.27",
      "value": 6.27,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/environment_tasks/claude-haiku-5-5/bart/ACCEPTANCE.json",
        "field": "/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 7424
      },
      "table2_marker": null
    },
    {
      "row_id": "a00081",
      "task": "temporal_discounting",
      "model": "jev_1_14",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "8",
      "n": 8,
      "value_raw": "0.5",
      "value": 0.5,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/environment_tasks/jev-1.14/temporal_discounting/ACCEPTANCE.json",
        "field": "/score_fraction"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 11
      },
      "table2_marker": null
    },
    {
      "row_id": "a00082",
      "task": "temporal_discounting",
      "model": "claude_opus_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00083",
      "task": "temporal_discounting",
      "model": "claude_sonnet_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00084",
      "task": "temporal_discounting",
      "model": "claude_haiku_5_5",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "9",
      "n": 9,
      "value_raw": "0.7777777777777778",
      "value": 0.7777777777777778,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/environment_tasks/claude-haiku-5-5/temporal_discounting/ACCEPTANCE.json",
        "field": "/score_fraction"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 12
      },
      "table2_marker": null
    },
    {
      "row_id": "a00085",
      "task": "two_step",
      "model": "jev_1_14",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "two_step",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "2000",
      "n": 2000,
      "value_raw": "0.547",
      "value": 0.547,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/environment_tasks/jev-1.14/two_step/ACCEPTANCE.json",
        "field": "/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 4000
      },
      "table2_marker": null
    },
    {
      "row_id": "a00086",
      "task": "two_step",
      "model": "claude_opus_5_5",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "two_step",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00087",
      "task": "two_step",
      "model": "claude_sonnet_5_5",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "two_step",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00088",
      "task": "two_step",
      "model": "claude_haiku_5_5",
      "protocol": "original_CogBench_v1_environment_with_new_typed_choice_adapter",
      "mode": "autonomous_generated_environment",
      "metric": "reward",
      "metric_variant": "mean_reward",
      "scope": "all",
      "unit": "task_specific_reward_units",
      "aggregation": "response_weighted_mean",
      "target": "autonomous_environment_reward",
      "mode_key": "two_step",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "2000",
      "n": 2000,
      "value_raw": "0.546",
      "value": 0.546,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/environment_tasks/claude-haiku-5-5/two_step/ACCEPTANCE.json",
        "field": "/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 4000
      },
      "table2_marker": null
    },
    {
      "row_id": "a00089",
      "task": "false_belief_story_qa",
      "model": "jev_1_14",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "false_belief_story_qa",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.7148815482148816",
      "value": 0.7148815482148816,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/false_belief_story_qa/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 5994
      },
      "table2_marker": null
    },
    {
      "row_id": "a00090",
      "task": "false_belief_story_qa",
      "model": "claude_opus_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "false_belief_story_qa",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00091",
      "task": "false_belief_story_qa",
      "model": "claude_sonnet_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "false_belief_story_qa",
      "value_status": "undefined",
      "status": "not_evaluated",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.e.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.e."
    },
    {
      "row_id": "a00092",
      "task": "false_belief_story_qa",
      "model": "claude_haiku_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "public_scenario_qa",
      "metric": "accuracy",
      "metric_variant": "original_answer_exact_match_mean",
      "scope": "all",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "original_objective_answer_or_optimal_action_label",
      "mode_key": "false_belief_story_qa",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "5994",
      "n": 5994,
      "value_raw": "0.8259926593259926",
      "value": 0.8259926593259926,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/claude-haiku-5-5/MODEL_ACCEPTANCE.json",
        "field": "/tasks/false_belief_story_qa/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 5994
      },
      "table2_marker": null
    },
    {
      "row_id": "a00093",
      "task": "politeness_pragmatics",
      "model": "jev_1_14",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "politeness_pragmatics",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/politeness_pragmatics/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 128
      },
      "table2_marker": null
    },
    {
      "row_id": "a00094",
      "task": "politeness_pragmatics",
      "model": "claude_opus_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "politeness_pragmatics",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00095",
      "task": "politeness_pragmatics",
      "model": "claude_sonnet_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "politeness_pragmatics",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00096",
      "task": "politeness_pragmatics",
      "model": "claude_haiku_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "politeness_pragmatics",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: stated probability. One of 128 responses is invalid because its probabilities sum to 0.67, so the acceptance record lists strict NLL as undefined. Two valid responses give zero probability to a human choice, so Table 3 prints +inf, which holds under any treatment of the invalid response.",
      "n_raw": "128",
      "n": 128,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/claude_verbalized_probability_20261030/RESULTS.json",
        "field": "/tasks/politeness_pragmatics/zero_probability_on_human_choice_items"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "stated probability",
        "model_calls": 128
      },
      "table2_marker": null
    },
    {
      "row_id": "a00097",
      "task": "pragmatic_language_interpretation",
      "model": "jev_1_14",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "pragmatic_language_interpretation",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/pragmatic_language_interpretation/score_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 170
      },
      "table2_marker": null
    },
    {
      "row_id": "a00098",
      "task": "pragmatic_language_interpretation",
      "model": "claude_opus_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "pragmatic_language_interpretation",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00099",
      "task": "pragmatic_language_interpretation",
      "model": "claude_sonnet_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "pragmatic_language_interpretation",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00100",
      "task": "pragmatic_language_interpretation",
      "model": "claude_haiku_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "aggregate_distribution_nll",
      "metric_variant": "human_response_weighted_mean_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "public_aggregate_human_distribution_not_individual_holdout_response",
      "mode_key": "pragmatic_language_interpretation",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: stated probability.",
      "n_raw": "170",
      "n": 170,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/claude_verbalized_probability_20261030/RESULTS.json",
        "field": "/tasks/pragmatic_language_interpretation/strict_nll"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "stated probability",
        "model_calls": 170
      },
      "table2_marker": null
    },
    {
      "row_id": "a00101",
      "task": "nonliteral_number_interpretation",
      "model": "jev_1_14",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "public_human_mean_slider_judgment",
      "mode_key": "nonliteral_number_interpretation",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "20.372090909090907",
      "value": 20.372090909090907,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/table2_followup/JEV_TWO_ADDITIONS_ACCEPTED_20261030.json",
        "field": "/models_or_tasks/nonliteral/summary/original_source_summary/absolute_error_to_human_mean_mean"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 330
      },
      "table2_marker": null
    },
    {
      "row_id": "a00102",
      "task": "nonliteral_number_interpretation",
      "model": "claude_opus_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "public_human_mean_slider_judgment",
      "mode_key": "nonliteral_number_interpretation",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Opus 5.5). Readout: generated FINAL line.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "15.335545454545455",
      "value": 15.335545454545455,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/paper_results_20261030/ACCEPTED_NONLITERAL_PAPER_VALUES_20261030.json",
        "field": "/models/claude-opus-5-5/primary_mae_to_human_mean_original_scale"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 330
      },
      "table2_marker": null
    },
    {
      "row_id": "a00103",
      "task": "nonliteral_number_interpretation",
      "model": "claude_sonnet_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "public_human_mean_slider_judgment",
      "mode_key": "nonliteral_number_interpretation",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Readout: generated FINAL line.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "20.807181818181817",
      "value": 20.807181818181817,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/paper_results_20261030/ACCEPTED_NONLITERAL_PAPER_VALUES_20261030.json",
        "field": "/models/claude-sonnet-5-5/primary_mae_to_human_mean_original_scale"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 330
      },
      "table2_marker": null
    },
    {
      "row_id": "a00104",
      "task": "nonliteral_number_interpretation",
      "model": "claude_haiku_5_5",
      "protocol": "social-public-full-materials-typed-v1",
      "mode": "aggregate_human_distribution_comparison",
      "metric": "mae",
      "metric_variant": "absolute_error_to_human_mean_mean",
      "scope": "all",
      "unit": "reported_percentage_points_0_to_100",
      "aggregation": "response_weighted_mean",
      "target": "public_human_mean_slider_judgment",
      "mode_key": "nonliteral_number_interpretation",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Claude Haiku 5.5). Readout: generated FINAL line.",
      "n_raw": "330",
      "n": 330,
      "value_raw": "22.521060606060605",
      "value": 22.521060606060605,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_results_20261029/paper_results_20261030/ACCEPTED_NONLITERAL_PAPER_VALUES_20261030.json",
        "field": "/models/claude-haiku-5-5/primary_mae_to_human_mean_original_scale"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": "generated FINAL line",
        "model_calls": 330
      },
      "table2_marker": null
    },
    {
      "row_id": "a00105",
      "task": "performance_self_assessment",
      "model": "jev_1_14",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "performance_self_assessment",
      "value_status": "defined_positive_infinity",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "Infinity",
      "value": "Infinity",
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_big_tasks_20261030/offline_prediction/jev-1.14/performance_self_assessment/ACCEPTANCE.json",
        "field": "/strict_nll_full_precision"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 108248
      },
      "table2_marker": null
    },
    {
      "row_id": "a00106",
      "task": "performance_self_assessment",
      "model": "claude_opus_5_5",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "performance_self_assessment",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Opus 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00107",
      "task": "performance_self_assessment",
      "model": "claude_sonnet_5_5",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "performance_self_assessment",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Sonnet 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00108",
      "task": "performance_self_assessment",
      "model": "claude_haiku_5_5",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "mean_typed_nll",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "performance_self_assessment",
      "value_status": "undefined",
      "status": "not_applicable_readout",
      "notes": "Table 3 API column (Claude Haiku 5.5). Table 3 prints n.a.",
      "n_raw": "",
      "n": null,
      "value_raw": "",
      "value": null,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": null,
        "field": null
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": null,
        "readout": null,
        "model_calls": null
      },
      "table2_marker": "n.a."
    },
    {
      "row_id": "a00109",
      "task": "multiple_cue_judgment",
      "model": "jev_1_14",
      "protocol": "original_full_prefix_typed",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll_clipped_1e12",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "multiple_cue_judgment",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice. Variant labelled \"source epsilon 1e-12\" in Table 3 records. Appendix value only. Table 3 prints the strict value, which is infinite.",
      "n_raw": "6136",
      "n": 6136,
      "value_raw": "3.069041860497198",
      "value": 3.069041860497198,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_small_tasks_20261030/offline_prompts/jev-1.14/MODEL_ACCEPTANCE.json",
        "field": "/tasks/multiple_cue_judgment/source_epsilon_nll/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 6136
      },
      "table2_marker": null
    },
    {
      "row_id": "a00110",
      "task": "performance_self_assessment",
      "model": "jev_1_14",
      "protocol": "full-original-prefix-integer-and-self-assessment-v1",
      "mode": "heldout_human_history_prediction",
      "metric": "human_nll",
      "metric_variant": "typed_choice_mean_nll_clipped_1e12",
      "scope": "all",
      "unit": "nats_per_response",
      "aggregation": "response_weighted_mean",
      "target": "observed_human_response",
      "mode_key": "performance_self_assessment",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 API column (Jev 1.14). Readout: native typed choice. Variant labelled \"source epsilon 1e-12\" in Table 3 records. Appendix value only. Table 3 prints the strict value, which is infinite.",
      "n_raw": "108248",
      "n": 108248,
      "value_raw": "2.7242904687571974",
      "value": 2.7242904687571974,
      "common_metric": true,
      "source": {
        "values_file": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
        "record": "runs/api_big_tasks_20261030/offline_prediction/jev-1.14/performance_self_assessment/ACCEPTANCE.json",
        "field": "/source_epsilon/value"
      },
      "runtime_context": {
        "device": "provider_hosted",
        "api_snapshot": "typesafe/jev-1.14-20261014",
        "readout": "native typed choice",
        "model_calls": 108248
      },
      "table2_marker": null
    },
    {
      "row_id": "t00001",
      "task": "temporal_discounting",
      "model": "laya",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 later-reward rate over realized reward questions: 0/6, with 3 payment questions excluded. Computed from saved choices with no new model call.",
      "n_raw": "6",
      "n": 6,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "record": "runs/probes_20261028/analysis/paper_body_revision_20261029/TABLE2_DIRECT_CONTROLS_20261029.json",
        "field": "/temporal/0/value"
      },
      "runtime_context": {
        "device": "cpu",
        "computed_from_saved_choices": true,
        "model_calls": 9
      },
      "table2_marker": null
    },
    {
      "row_id": "t00002",
      "task": "temporal_discounting",
      "model": "intern_decision_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 later-reward rate over realized reward questions: 1/6, with 3 payment questions excluded. Computed from saved choices with no new model call.",
      "n_raw": "6",
      "n": 6,
      "value_raw": "0.16666666666666666",
      "value": 0.16666666666666666,
      "common_metric": true,
      "source": {
        "record": "runs/probes_20261028/analysis/paper_body_revision_20261029/TABLE2_DIRECT_CONTROLS_20261029.json",
        "field": "/temporal/1/value"
      },
      "runtime_context": {
        "device": "cpu",
        "computed_from_saved_choices": true,
        "model_calls": 9
      },
      "table2_marker": null
    },
    {
      "row_id": "t00003",
      "task": "temporal_discounting",
      "model": "llama8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 later-reward rate over realized reward questions: 0/6, with 3 payment questions excluded. Computed from saved choices with no new model call.",
      "n_raw": "6",
      "n": 6,
      "value_raw": "0.0",
      "value": 0.0,
      "common_metric": true,
      "source": {
        "record": "runs/probes_20261028/analysis/paper_body_revision_20261029/TABLE2_DIRECT_CONTROLS_20261029.json",
        "field": "/temporal/2/value"
      },
      "runtime_context": {
        "device": "cpu",
        "computed_from_saved_choices": true,
        "model_calls": 9
      },
      "table2_marker": null
    },
    {
      "row_id": "t00004",
      "task": "temporal_discounting",
      "model": "minitaur8b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 later-reward rate over realized reward questions: 8/10, with 3 payment questions excluded. Computed from saved choices with no new model call.",
      "n_raw": "10",
      "n": 10,
      "value_raw": "0.8",
      "value": 0.8,
      "common_metric": true,
      "source": {
        "record": "runs/probes_20261028/analysis/paper_body_revision_20261029/TABLE2_DIRECT_CONTROLS_20261029.json",
        "field": "/temporal/3/value"
      },
      "runtime_context": {
        "device": "cpu",
        "computed_from_saved_choices": true,
        "model_calls": 13
      },
      "table2_marker": null
    },
    {
      "row_id": "t00005",
      "task": "temporal_discounting",
      "model": "qwen35_4b",
      "protocol": "original_environment",
      "mode": "autonomous_generated_environment",
      "metric": "preference",
      "metric_variant": "delayed_reward_choice_rate",
      "scope": "reward_choice_questions",
      "unit": "fraction",
      "aggregation": "response_weighted_mean",
      "target": "later_receipt_among_reward_choice_questions",
      "mode_key": "temporal_discounting",
      "value_status": "finite",
      "status": "available_full_protocol",
      "notes": "Table 3 later-reward rate over realized reward questions: 9/10, with 3 payment questions excluded. Computed from saved choices with no new model call.",
      "n_raw": "10",
      "n": 10,
      "value_raw": "0.9",
      "value": 0.9,
      "common_metric": true,
      "source": {
        "record": "runs/probes_20261028/analysis/paper_body_revision_20261029/TABLE2_DIRECT_CONTROLS_20261029.json",
        "field": "/temporal/4/value"
      },
      "runtime_context": {
        "device": "cpu",
        "computed_from_saved_choices": true,
        "model_calls": 13
      },
      "table2_marker": null
    }
  ],
  "model_access": {
    "laya": "open_weights",
    "intern_decision_4b": "open_weights",
    "llama8b": "open_weights",
    "minitaur8b": "open_weights",
    "qwen35_4b": "open_weights",
    "gaussian_ts": "classical_policy",
    "jev_1_14": "api",
    "claude_opus_5_5": "api",
    "claude_sonnet_5_5": "api",
    "claude_haiku_5_5": "api"
  },
  "api_extension": {
    "source_acceptance": "table2_values_20261030.json (main results table, paper commit 5a0f491)",
    "accepted_utc": "2026-10-30T09:08:57.534198+00:00",
    "rows_added": 110,
    "models": {
      "jev_1_14": {
        "label": "Jev 1.14",
        "requested": "typesafe/jev-1.14",
        "snapshot": "typesafe/jev-1.14-20261014",
        "route": "OpenRouter decisions endpoint"
      },
      "claude_opus_5_5": {
        "label": "Claude Opus 5.5",
        "requested": "claude-opus-5-5",
        "route": "Anthropic Messages API"
      },
      "claude_sonnet_5_5": {
        "label": "Claude Sonnet 5.5",
        "requested": "claude-sonnet-5-5",
        "route": "Anthropic Messages API"
      },
      "claude_haiku_5_5": {
        "label": "Claude Haiku 5.5",
        "requested": "claude-haiku-5-5",
        "route": "Anthropic Messages API"
      }
    },
    "markers": {
      "n.a.": "not_applicable_readout",
      "part.": "partial_records_retained",
      "n.e.": "not_evaluated",
      "n.f.": "non_full_prefix_only",
      "canc.": "cancelled_before_generation",
      "inv.": "all_responses_invalid"
    }
  },
  "table2_temporal_rows": {
    "source": "runs/probes_20261028/analysis/paper_body_revision_20261029/TABLE2_DIRECT_CONTROLS_20261029.json",
    "rows_added": 5
  }
}