Dengzhe Hou

Dengzhe Hou

Assistant Professor @ Tohoku GSIS | JSPS DC2

Sitemap

A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.

Pages

Posts

portfolio

publications

Commonality of neuronal coherence for motor skill acquisition and interlimb transfer

Published in Scientific Reports, 2025

This study identified common neural coherence patterns underlying both motor skill acquisition and interlimb transfer, suggesting shared neurophysiological mechanisms for these two motor learning processes.

Recommended citation: Zhao, J., Wang, Y., Hou, D., Sun, S., Négyesi, J., Inada, H., Shioiri, S., & Nagatomi, R. (2025). Commonality of neuronal coherence for motor skill acquisition and interlimb transfer. Scientific Reports, 15(1), 26276.
Download Paper

EEG Activity Over Ipsilateral and Contralateral M1 During Simple and Complex Hand Tasks: Variations with Motor Learning

Published in Frontiers in Neuroscience, 2025

This study characterized EEG activity over ipsilateral and contralateral primary motor cortex during simple and complex hand tasks, revealing how motor cortex lateralization changes across motor learning.

Recommended citation: Zhao, J., Wang, Y., Hou, D., Négyesi, J., Qiu, D. L., & Nagatomi, R. (2025). EEG activity over ipsilateral and contralateral M1 during simple and complex hand tasks: Variations with motor learning. Frontiers in Neuroscience, 19, 1681250.
Download Paper

Task-constrained self-initiated attention shifts are indexed by frontal-midline theta ramping

Published in Frontiers in Human Neuroscience, 2025

We investigated the EEG signatures of voluntary, self-initiated attention shifts during visual search. Frontal-midline theta oscillations showed a characteristic ramping pattern prior to attention shifts, reflecting the cognitive demands of self-initiated attentional control.

Recommended citation: Hou, D., Sun, S., Hatori, Y., Tseng, C., & Shioiri, S. (2025). Task-constrained self-initiated attention shifts are indexed by frontal-midline theta ramping. Frontiers in Human Neuroscience.
Download Paper

Physics-Aware Video Instance Removal Benchmark

Published in CVPR 2026 Workshop on Video Generation and Beyond Evaluation (VGBE), 2026

We introduce PVIR, a benchmark of 95 videos with a decoupled human evaluation protocol that reveals current video instance removal methods still treat object erasure as 2D texture filling rather than physics-aware scene reconstruction, particularly failing on complex physical side effects like reflections and shadows.

Recommended citation: Li, Z., Chen, X., Jiang, L., Hou, D., Lin, F., Yamada, K., Gao, X., & Tu, Z. (2026). Physics-Aware Video Instance Removal Benchmark. CVPR 2026 Workshop VGBE.
Download Paper

WMF-AM: Probing LLM Working Memory via Depth-Parameterized Cumulative State Tracking

Published in arXiv preprint, 2026

We introduce Working Memory Fidelity-Active Manipulation (WMF-AM), a probe of cumulative state tracking that isolates within-pass cumulative load by parameterizing depth K. Testing 20 open-weight models (0.5B–35B) across 13 families, our probe predicts agent performance with r = 0.612 (p < 0.001).

Recommended citation: Hou, D., Jiang, L., Li, D., Li, Z., Lin, F., & Yamada, K. D. (2026). WMF-AM: Probing LLM Working Memory via Depth-Parameterized Cumulative State Tracking. arXiv preprint arXiv:2603.27343.
Download Paper

Vibe Medicine: Redefining Biomedical Research Through Human-AI Co-Work

Published in arXiv preprint (under review at Meta-Radiology), 2026

A framework where clinicians and researchers direct skill-augmented AI agents through natural language to execute complex biomedical workflows. Case studies span rare disease diagnosis, drug repurposing, and clinical trial design across a curated library of 1,000+ medical skills.

Recommended citation: Wu, Z., Xu, S., Chen, B., Wan, S., Li, Y., Ruan, W., Lyu, Y., Li, S., Zhu, D., Liu, T., & Zhao, L. (2026). Vibe Medicine: Redefining Biomedical Research Through Human-AI Co-Work. arXiv preprint arXiv:2604.23674.
Download Paper

Same Brain, Different Prediction: How Preprocessing Choices Undermine EEG Decoding Reliability

Published in arXiv preprint, 2026

Across six datasets and four paradigms, up to 42% of trial-level EEG predictions flip when only the preprocessing pipeline changes. We introduce Walsh-Hadamard decomposition to characterize sensitivity, Preprocessing Uncertainty as a diagnostic, and Normalized Adaptive PGI as a regularization fix.

Recommended citation: Hou, D., Wu, Z., Jiang, L., Li, Z., Lin, F., & Yamada, K. D. (2026). Same Brain, Different Prediction: How Preprocessing Choices Undermine EEG Decoding Reliability. arXiv preprint arXiv:2605.07212.
Download Paper

Subject-Specific Analysis of Self-Initiated Attention Shifts from EEG with Controlled Internal and External Attention Conditions

Published in IEEE International Conference on Systems, Man, and Cybernetics (SMC 2026), Bellevue, USA, 2026

We analyze self-initiated attention shifts in EEG using subject-level machine learning, combining frequency-specific topographic patterns with SHAP feature attribution. Higher-frequency bands and frontal regions contribute most to within-subject classification.

Recommended citation: Zeng, Y., Hou, D., Zhang, Z., Sun, S., Huang, Y., Tseng, C., & Shioiri, S. (2026). Subject-Specific Analysis of Self-Initiated Attention Shifts from EEG with Controlled Internal and External Attention Conditions. IEEE International Conference on Systems, Man, and Cybernetics (SMC 2026).
Download Paper

PathCal: State-Aware Reflection-Marker Calibration for Efficient Reasoning

Published in arXiv preprint (under review, NeurIPS 2026), 2026

Reasoning models spend tokens re-checking themselves. PathCal calibrates when reflection markers fire, conditioned on the reasoning state, to cut that overhead without losing accuracy.

Recommended citation: Jiang, L., Li, Z., Xing, S., Li, P., Takahashi, T., Hou, D., Tu, Z., Yamada, K., & Lin, F. (2026). PathCal: State-Aware Reflection-Marker Calibration for Efficient Reasoning. arXiv preprint arXiv:2605.23074.
Download Paper

Self-Evolving Agent Engineering for Healthcare: Methodologies and Applications

Published in Preprints.org, 2026

A survey of how healthcare AI agents are built to improve themselves, covering the methodologies behind self-evolving agent engineering and where they are actually deployed in clinical and biomedical settings.

Recommended citation: Hou, D., Wu, Z., Zeng, Y., Jiang, L., Lin, F., & Yamada, K. (2026). Self-Evolving Agent Engineering for Healthcare: Methodologies and Applications. Preprints.org.
Download Paper

A Compact Kolmogorov–Arnold Network Mixer for Long-Term Time Series Forecasting

Published in Scientific Reports, 2026

Can a Kolmogorov–Arnold Network serve as the modeling core for forecasting, rather than a decoration on an existing backbone? KANMixer is a deliberately compact architecture built to answer that, and the answer turns out to be conditional.

Recommended citation: Jiang, L., Hou, D., Wang, Y., Su, Y., Xing, S., Chen, W., Zhang, X., Tu, Z., Zhang, Z., Lin, F., Zielewski, M., & Yamada, K. (2026). A compact Kolmogorov–Arnold network mixer for long-term time series forecasting. Scientific Reports. https://doi.org/10.1038/s41598-026-59667-5
Download Paper

TimePre: Bridging Accuracy, Efficiency, and Stability in Probabilistic Time-Series Forecasting

Published in Transactions on Machine Learning Research (TMLR), accepted, 2026

Probabilistic forecasters usually trade one of accuracy, speed, or training stability for the other two. TimePre keeps all three, and holds up across standard long-horizon benchmarks.

Recommended citation: Jiang, L., Xu, L., Li, P., Hou, D., Ge, Q., Zhuang, D., Xing, S., Chen, W., Gao, X., et al. (2026). TimePre: Bridging Accuracy, Efficiency, and Stability in Probabilistic Time-Series Forecasting. Transactions on Machine Learning Research.
Download Paper

CogArena: A Multimethod Evaluation of Cognitive Ability Structure in Large Language Models

Published in arXiv preprint (under review, AAAI 2027), 2026

Rather than asking how well a model scores, CogArena asks how its abilities are organized: a multimethod battery that recovers the latent structure of cognitive ability in large language models.

Recommended citation: Hou, D., Jiang, L., Lin, F., & Yamada, K. D. (2026). CogArena: A Multimethod Evaluation of Cognitive Ability Structure in Large Language Models. arXiv preprint arXiv:2607.24999.
Download Paper

CogEEGAgent: Toward Autonomous Cognitive EEG Analysis with Grounded Execution and Selection-Aware Verification

Published in arXiv preprint (under review, AAAI 2027), 2026

An LLM agent that runs cognitive EEG analyses end to end, grounding every step in executable code and verifying its own analytic choices instead of reporting whichever pipeline happened to look best.

Recommended citation: Hou, D., Jiang, L., Lin, F., & Yamada, K. D. (2026). CogEEGAgent: Toward Autonomous Cognitive EEG Analysis with Grounded Execution and Selection-Aware Verification. arXiv preprint arXiv:2607.25045.
Download Paper

Control-Diverse Reinforcement Fine-Tuning: Decoupling the Shared Control Bottleneck of RL Post-Training

Published in arXiv preprint (under review, AAAI 2027), 2026

RL post-training is usually explained by which circuits it activates. We separate activation from control, and find that control concentrates on a shared set of components across tasks even when activations look diverse.

Recommended citation: Tan, B., Wang, J., Hou, D., Jiang, L., Wu, Z., Shen, Y., Lin, F., Yamada, K., & Koike, A. (2026). Control-Diverse Reinforcement Fine-Tuning: Decoupling the Shared Control Bottleneck of RL Post-Training. arXiv preprint arXiv:2608.08224.
Download Paper

talks

Self-initiation of attentional shift during visual search

Published:

Oral presentation at the European Conference on Visual Perception (ECVP) 2023. This talk presented EEG and eye-tracking evidence for the neural mechanisms underlying self-initiated attentional shifts during free visual search. We characterized the temporal dynamics of oscillatory activity preceding voluntary attention shifts and discussed implications for models of endogenous attentional control.

Exploring Mechanisms of Self-initiated Attention Shifts: Analysis of Theta Wave

Published:

Poster presentation at the 16th Asia Pacific Conference on Vision (APCV 2024). This work presents a theta oscillation analysis of EEG data recorded during self-initiated attention shifts in a visual search task. Frontal-midline theta power showed systematic ramping prior to voluntary saccades, linking preparatory neural dynamics to the initiation of attention shifts.

EEG-Based Decoding of Voluntary Attentional Control: Differentiating Self-Initiated Shifts in a Visual Search Paradigm

Published:

Poster presentation at the joint EPC (European Conference on Eye Movements) and APCV (Asia Pacific Conference on Vision) 2025 meeting. We demonstrated EEG-based multivariate decoding to differentiate self-initiated attention shifts from externally driven shifts in a visual search paradigm, revealing distinct temporal dynamics in voluntary attentional control.

Neurotechnology

Published:

Invited talk in Open Forum OF7, “テクノロジーの質的進化と組織統制” (Qualitative Evolution of Technology and Organizational Control), at the 44th Annual Conference of the Robotics Society of Japan, Kanazawa University Kakuma Campus.

teaching

Teaching Assistant

Graduate & Undergraduate Courses, Graduate School of Information Sciences, Tohoku University, 2022

Teaching Assistant at the Graduate School of Information Sciences, Tohoku University (April 2022 – March 2025).