From Conflict to Consensus: Boosting Medical Reasoning via Multi-Round Agentic RAG

TL;DR

This paper introduces MA-RAG, a multi-round agentic retrieval framework that enhances medical reasoning in LLMs by iterative evidence refinement, significantly improving accuracy in medical question-answering tasks.

Contribution

The paper presents MA-RAG, a novel multi-round reasoning and retrieval method that extends self-consistency principles for improved medical question-answering accuracy.

Findings

MA-RAG outperforms baseline models by +6.8 accuracy points on average.

It effectively transforms conflicting responses into targeted queries for better evidence retrieval.

The framework demonstrates consistent improvements across 7 medical Q&A benchmarks.

Abstract

Large Language Models (LLMs) exhibit high reasoning capacity in medical question-answering, but their tendency to produce hallucinations and outdated knowledge poses critical risks in healthcare fields. While Retrieval-Augmented Generation (RAG) mitigates these issues, existing methods rely on noisy token-level signals and lack the multi-round refinement required for complex reasoning. In the paper, we propose MA-RAG (Multi-Round Agentic RAG), a framework that facilitates test-time scaling for complex medical reasoning by iteratively evolving both external evidence and internal reasoning history within an agentic refinement loop. At each round, the agent transforms semantic conflict among candidate responses into actionable queries to retrieve external evidence, while optimizing history reasoning traces to mitigate long-context degradation. MA-RAG extends the self-consistency principle by leveraging the lack of consistency as a proactive signal for multi-round agentic reasoning and retrieval, and mirrors a boosting mechanism that iteratively minimizes the residual error toward a stable, high-fidelity medical consensus. Extensive evaluations across 7 medical Q&A benchmarks show that MA-RAG consistently surpasses competitive inference-time scaling and RAG baselines, delivering substantial +6.8 points on average accuracy over the backbone model. Our code is available at https://github.com/NJU-RL/MA-RAG.