Healthcare AI GYM for Medical Agents

• URL: http://arxiv.org/abs/2605.02943v1
• Date: Fri, 01 May 2026 10:11:22 GMT
• Title: Healthcare AI GYM for Medical Agents
• Abstract summary: We present a comprehensive empirical study of multi-turn agentic RL for medical AI, built on gym, a gymnasium-compatible environment.<n>Our analysis reveals that agentic multi-turn structure degrades into verbose single-turn monologues.<n>To improve training efficiency and stability, we propose Turn-level Truncated On-Policy Distillation (TT-OPD)
• Score: 2.3677513428867614
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Clinical reasoning demands multi-step interactions -- gathering patient history, ordering tests, interpreting results, and making safe treatment decisions -- yet a unified training environment provides the breadth of clinical domains and specialized tools to train generalizable medical AI agents through reinforcement learning remains elusive. We present a comprehensive empirical study of multi-turn agentic RL for medical AI, built on \gym{}, a gymnasium-compatible environment spanning 10 clinical domains with 3.6K+ tasks, 135 domain-specific tools, and a knowledge base of 828K medical passages. Our analysis reveals that agentic multi-turn structure degrades into verbose single-turn monologues, characterized by monotonic length explosion and a simultaneous erosion of tool-use frequency. We characterize how this collapse, alongside distillation instability, stems from the misalignment of sparse terminal rewards with sequential clinical trajectories. We find that vanilla GRPO achieves strong final accuracy on some benchmarks but suffers from training instability, evidenced by significant oscillations in response length and prolonged convergence periods. To improve training efficiency and stability, we propose Turn-level Truncated On-Policy Distillation (TT-OPD), a self-distillation framework where a gradient-free EMA teacher leverages outcome-privileged information to provide dense, outcome-aware KL regularization at every conversation turn. TT-OPD achieves the best performance on 10 of 18 benchmarks with an average +3.9~pp improvement over the non-RL baseline with faster early convergence, controlled response length, and sustained multi-turn tool use.

Related papers

• Asclepius: An Adaptive Harness for Long-Horizon Clinical Agents [10.003509950132251]
  Long-horizon execution failures manifest measurably in a single rollout under structured, multi-dimensional grading.<n>We introduce Asclepius, an adaptive agent scaffolding with a self-evolving harness.<n>Asclepius improves critical-action correctness by 22% over a strong baseline agent framework.
  arXiv  Detail & Related papers  (2026-09-11T21:14:59Z)
• ResidencyRL: Reinforcement Learning in Simulated Clinical Environments [53.76751756107077]
  We present ResidencyRL, a reinforcement learning (RL) method for training clinical artificial intelligence (AI) agents through simulated multi-turn clinical encounters.<n>On heldout evaluations, the ResidencyRL agent improves diagnostic accuracy by 7.0% under adversarial conditions.<n>Blinded expert clinicians validated these gains, preferring the trained agent in 87.6% of side-by-side comparisons.
  arXiv  Detail & Related papers  (2026-08-07T17:04:41Z)
• MA-RAG: Multi-Agent Retrieval-Augmented Generation for Query-Driven Summarization of Longitudinal Parkinson's Disease Assessments [0.8859671910259244]
  MA-RAG is a query-driven multi-agent retrieval-augmented generation framework.<n>It decomposes clinical reasoning into domain-specialized agents.<n>It produces factually correct and temporally consistent responses for structured longitudinal assessment data.
  arXiv  Detail & Related papers  (2026-07-30T23:20:00Z)
• MedCTA: A Benchmark for Clinical Tool Agents [47.47354499871572]
  We introduce MedCTA, a benchmark for evaluating medical tool agents on clinician-validated, step-implicit tasks.<n>MedCTA comprises 107 real-world clinical tasks with clinician-verified executable trajectories over 5 deployed tools.<n>We benchmark 18 open- and closed-source multimodal models and find that even frontier systems remain brittle in multi-step clinical tool use.
  arXiv  Detail & Related papers  (2026-06-10T06:26:52Z)
• Benchmarking Multi-turn Medical Diagnosis: Hold, Lure, and Self-Correction [72.89352076103889]
  Large language models (LLMs) achieve high accuracy in medical diagnosis when all clinical information is provided in a single turn.<n>We introduce MINT, a high-fidelity, multi-turn medical diagnosis benchmark comprising 1,035 cases.<n>We uncover three persistent behavioral patterns that significantly impact diagnostic decisions.
  arXiv  Detail & Related papers  (2026-04-06T00:23:10Z)
• Cerebra: A Multidisciplinary AI Board for Multimodal Dementia Characterization and Risk Assessment [56.62016795093786]
  Cerebra is an interactive multi-agent AI team that coordinates specialized agents for EHR, clinical notes, and medical imaging analysis.<n>It supports privacy-preserving deployment by operating on structured representations and remains robust when modalities are incomplete.<n>Cerebra significantly improved expert performance, increasing accuracy by 17.5 percentage points in prospective dementia risk estimation.
  arXiv  Detail & Related papers  (2026-03-23T05:46:45Z)
• A Multi-Agent Framework for Interpreting Multivariate Physiological Time Series [9.72130666902599]
  We present Vivaldi, a role-structured multi-agent system that explains multivariate physiological time series.<n>Our experiments show that agentic pipelines substantially benefit non-thinking and medically fine-tuned models.<n>We find that explicit tool-based computation is decisive for codifiable clinical metrics, whereas subjective targets, such as pain scores and length of stay, show limited or inconsistent changes.
  arXiv  Detail & Related papers  (2026-03-04T14:55:46Z)
• MMedExpert-R1: Strengthening Multimodal Medical Reasoning via Domain-Specific Adaptation and Clinical Guideline Reinforcement [63.82954136824963]
  Medical Vision-Language Models excel at perception tasks with complex clinical reasoning required in real-world scenarios.<n>We propose a novel reasoning MedVLM that addresses these challenges through domain-specific adaptation and guideline reinforcement.
  arXiv  Detail & Related papers  (2026-01-16T02:32:07Z)
• MedKGent: A Large Language Model Agent Framework for Constructing Temporally Evolving Medical Knowledge Graph [57.54231831309079]
  We introduce MedKGent, a framework for constructing temporally evolving medical Knowledge Graphs.<n>We simulate the emergence of biomedical knowledge via a fine-grained daily time series.<n>The resulting KG contains 156,275 entities and 2,971,384 relational triples.
  arXiv  Detail & Related papers  (2025-08-17T15:14:03Z)
• Cog-TiPRO: Iterative Prompt Refinement with LLMs to Detect Cognitive Decline via Longitudinal Voice Assistant Commands [15.594650981318514]
  Early detection of cognitive decline is crucial for enabling interventions that can slow neurodegenerative disease progression.<n>Our pilot study investigates voice assistant systems (VAS) as non-invasive tools for detecting cognitive decline through longitudinal analysis of speech patterns in voice commands.
  arXiv  Detail & Related papers  (2025-05-22T05:40:12Z)
• AgentClinic: a multimodal agent benchmark to evaluate AI in simulated clinical environments [2.567146936147657]
  We introduce AgentClinic, a multimodal agent benchmark for evaluating large language models (LLM) in simulated clinical environments.<n>We find that solving MedQA problems in the sequential decision-making format of AgentClinic is considerably more challenging, resulting in diagnostic accuracies that can drop to below a tenth of the original accuracy.
  arXiv  Detail & Related papers  (2024-05-13T17:38:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.