A Survey on Recursive Self-Improvementin Healthcare AI Agents

Dengzhe Houa,b,*Zihao WucYuwen ZengdLingyu JiangaFangzhou Lina,e,fKazunori D Yamadaa,b

Author affiliations

a Graduate School of Information Sciences, Tohoku University
b Unprecedented-scale Data Analytics Center, Tohoku University
c School of Computing, University of Georgia
d Advanced Institute of Convergence Knowledge Informatics, Tohoku University
e Texas A&M University
f Worcester Polytechnic Institute

Twenty-eight selected healthcare-agent studies by update target and public month
Timeline of selected healthcare-agent studies, grouped by update target and positioned by first public release. Shaded ribbons illustrate conceptual connections between update targets. Year intervals are not drawn to scale. Explore study-level evidence ↗

Abstract

Advances in persistent memory, reusable skills and adaptive training are making recursive self-improvement (RSI) increasingly relevant to healthcare agents.

RSI concerns changes to the procedures that generate future improvements. This survey connects these developments with healthcare evidence by distinguishing component updates, training-generator changes and revisions of improvement procedures. We examine applications in diagnosis, reporting and question answering, together with challenges in evaluation, reliability, privacy and update governance. Current evidence supports persistent component updates and selected generator gains, while the effects of revised improvement procedures on subsequent learning remain untested in the mapped studies. An accompanying evidence map and evaluation templates support study comparison and future research.

Survey Contents

Select a chapter to show its sections. Hover a chapter to preview its figure. Section links open the corresponding mapped studies.

Mechanism Examples

Six published frameworks illustrate what feedback changes. Select a study to inspect its mechanism and source record. Examples advance every 3 seconds. Hover to pause.

Full six-panel figure ↗  ·  Sources and adaptation notes ↗

Recently Added

We maintain this collection as new healthcare-agent studies appear. Papers first made available after 4 November 2026 will be listed here after review.

No papers have been added after the survey cutoff yet. The 85-study survey snapshot remains available as version 0.3.2.

Help improve the collection by suggesting a paper, reporting a correction or asking about scope. See the contribution guide for what to include.

News and Updates

  1. Preprints.org v2 is now public. The revised survey is also available on alphaXiv. The fixed evidence release remains v0.3.2.

  2. The revised survey is now available on alphaXiv and has been submitted to Meta-Radiology. The 85-study evidence map and evaluation resources accompany the paper.

  3. The connected paper map, paper collection and evidence resources are available. The fixed evidence release remains v0.3.2.

  4. Unified search cutoff across arXiv, PubMed, Europe PMC and OpenAlex.

Read the revised preprint on Preprints.org (v2) or alphaXiv. The first version was posted on 22 May 2026 under an earlier title.

BibTeX

Cite the survey and the fixed resource version when reusing its records.