{"ID":2869465,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2509.15154","arxiv_id":"2509.15154","title":"MedFact-R1: Towards Factual Medical Reasoning via Pseudo-Label Augmentation","abstract":"Ensuring factual consistency and reliable reasoning remains a critical challenge for medical vision-language models. We introduce MEDFACT-R1, a two-stage framework that integrates external knowledge grounding with reinforcement learning to improve the factual medical reasoning. The first stage uses pseudo-label supervised fine-tuning (SFT) to incorporate external factual expertise; while the second stage applies Group Relative Policy Optimization (GRPO) with four tailored factual reward signals to encourage self-consistent reasoning. Across three public medical QA benchmarks, MEDFACT-R1 delivers up to 22.5% absolute improvement in factual accuracy over previous state-of-the-art methods. Ablation studies highlight the necessity of pseudo-label SFT cold start and validate the contribution of each GRPO reward, underscoring the synergy between knowledge grounding and RL-driven reasoning for trustworthy medical AI. Codes are released at https://github.com/Garfieldgengliang/MEDFACT-R1.","short_abstract":"Ensuring factual consistency and reliable reasoning remains a critical challenge for medical vision-language models. We introduce MEDFACT-R1, a two-stage framework that integrates external knowledge grounding with reinforcement learning to improve the factual medical reasoning. The first stage uses pseudo-label supervi...","url_abs":"https://arxiv.org/abs/2509.15154","url_pdf":"https://arxiv.org/pdf/2509.15154v1","authors":"[\"Gengliang Li\",\"Rongyu Chen\",\"Bin Li\",\"Linlin Yang\",\"Guodong Ding\"]","published":"2025-09-18T16:59:59Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"Reinforcement Learning\",\"Language Model\"]","has_code":false,"code_links":[{"ID":609688,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2869465,"paper_url":"https://arxiv.org/abs/2509.15154","paper_title":"MedFact-R1: Towards Factual Medical Reasoning via Pseudo-Label Augmentation","repo_url":"https://github.com/Garfieldgengliang/MEDFACT-R1","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}