Reinforcement Learning for Evidence-Seeking Diagnostic Reasoning with Large Language Models
For clinical AI, this work addresses the gap between passive LLM inference and the iterative evidence-seeking nature of real-world diagnosis, enabling more autonomous diagnostic assistants.
The paper formalizes medical diagnosis as an iterative evidence-seeking task and uses reinforcement learning with verifiable rewards to train LLMs to actively gather diagnostic information. The resulting model achieves performance comparable to larger baselines, and the proposed examination simulator generates more biologically plausible clinical feedback than vanilla LLMs.
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.