Research map / Large language models / Retrieval and knowledge
From Conflict to Consensus: Boosting Medical Reasoning via Multi-Round Agentic RAG
Wenhao Wu, Zhentao Tang, Yafu Li, Shixiong Kai, Mingxuan Yuan, Zhenhong Sun, Chunlin Chen, Zhi Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add multi-round self-refinement (Multi-Refine) (ablation) | Qwen3-8B · 7 medical benchmarks avg | accuracy: 55.4 -> 58.7 | Improved |
| add conflict-guided retrieval agent (ablation) | MA-RAG on Qwen3-8B · 7 medical benchmarks avg | accuracy: 58.7 -> 60.6 | Improved |
| add ranking agent for history context (ablation) | MA-RAG on Qwen3-8B · 7 medical benchmarks avg | accuracy: 60.6 -> 62.2 | Improved |
| use BERT verifier score instead of entropy in ranking agent (ablation) | MA-RAG · 7 medical benchmarks avg | accuracy: 61.0 -> 62.2 | Improved |
| use self-consistency (single-round voting) | Qwen3-8B · MedXpertQA | accuracy: 16.1 -> 15.7 | No clear effect |
| naive single-round RAG | Qwen3-8B · MedQA | accuracy: 71.1 -> 69.9 | Hurt |
| increase max refinement rounds from 1 to 2 (ablation) | MA-RAG-ext · 7 medical benchmarks avg | accuracy: 57.5 -> 61.5 | Improved |
| increase max refinement rounds from 4 to 8 (ablation) | MA-RAG-ext · 7 medical benchmarks avg | accuracy: 62.1 -> 62.2 | No clear effect |
| increase candidate pool size from 2 to 8 (ablation) | MA-RAG-ext · 7 medical benchmarks avg | accuracy: 58.5 -> 62.2 | Improved |
| scale backbone to Qwen3-32B | MA-RAG · 7 medical benchmarks avg | accuracy: 64.7 -> 70.2 | Improved |
| add retrieval agent on larger backbone (ablation) | MA-RAG on Qwen3-32B · MedXpertQA | accuracy: 21.2 -> 24.9 | Improved |
| MA-RAG vs MDAgents on inference cost | MA-RAG-ext · MedXpertQA | wall-clock seconds: 70.7 vs 146.2 (MDAgents) | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MDAgents — MDAgents: An Adaptive Collaboration of LLMs for Medical Decision-Making (NeurIPS 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MedLA: A Logic-Driven Multi-Agent Framework for Complex Medical Reasoning with Large Language Models (AAAI 2026 · similarity 0.89)
- MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains (ICLR 2026 · similarity 0.88)
- Simulating Viva Voce Examinations to Evaluate Clinical Reasoning in Large Language Models (NeurIPS 2025 · similarity 0.88)
- MTBBench: A Multimodal Sequential Clinical Decision-Making Benchmark in Oncology (NeurIPS 2025 · similarity 0.88)
- KGARevion: An AI Agent for Knowledge-Intensive Biomedical QA (ICLR 2025 · similarity 0.88)
- Chiron-o1: Igniting Multimodal Large Language Models towards Generalizable Medical Reasoning via Mentor-Intern Collaborative Search (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
