Reinforcement Learning for Evidence-Seeking Diagnostic Reasoning with Large Language Models
Quick Answer
This study introduces a framework that formalizes medical diagnosis as an Iterative Evidence-Seeking Task using Reinforcement Learning with Verifiable Rewards (RLVR) and a novel Examination Simulator (RAGES).
Quick Take
This study introduces a framework that formalizes medical diagnosis as an Iterative Evidence-Seeking Task using Reinforcement Learning with Verifiable Rewards (RLVR) and a novel Examination Simulator (RAGES). The approach enables Large Language Models (LLMs) to evolve from passive responders to autonomous diagnostic assistants, achieving performance comparable to larger models while generating biologically plausible clinical feedback.
Key Points
- Introduces RLVR to enhance reasoning in medical diagnostics.
- RAGES provides realistic, knowledge-grounded follow-up evidence.
- Empirical results show LLMs transitioning to autonomous assistants.
- Framework achieves comparable performance to larger reasoning models.
- RAGES outperforms vanilla LLMs in generating clinical feedback.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Recent reasoning-centric Large Language Models (LLMs) have made significant strides, yet they predominantly operate on a passive-inference pattern that assumes complete information. In contrast, real-world clinical intelligence is inherently an iterative investigative process requiring strategic evidence acquisition. To bridge this gap, we formalize medical diagnosis as an Iterative Evidence-Seeking Task. We leverage Reinforcement Learning with Verifiable Rewards (RLVR) to elicit intrinsic reasoning within a closed-loop environment, guided by a novel suite of rewards that enforce diagnostic precision and examination consistency. To facilitate this, we introduce the Retrieval-Augmented Generation-based Examination Simulator (RAGES), a high-fidelity clinical oracle that provides realistic, knowledge-grounded follow-up evidence. Empirical results across diverse datasets demonstrate that our framework enables LLMs to transition from passive responders to autonomous assistants. Notably, our model demonstrates comparable performance to larger and reasoning-enhanced baselines, while RAGES proves superior to vanilla LLMs in generating biologically plausible clinical feedback.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.02983 [cs.AI] |
| (or arXiv:2607.02983v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.02983 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Shengyi Hua [view email]
[v1]
Fri, 3 Jul 2026 05:43:00 UTC (4,396 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Procedural Memory Distillation: Online Reflection for Self-Improving Language Models
Procedural Memory Distillation (PMD) enhances reinforcement learning by converting cross-episode signals into reusable memory, improving Qwen3-8B and OLMo3-Instruct-7B models by 3.8-5.5% on SCIKNOWEVAL and 7.9-13.6% on . The co-evolution of policy and memory allows for more effective self-supervision, demonstrating significant performance gains when both components are active.