Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.15154 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918143745589248 |
|---|---|
| author | Li, Gengliang Chen, Rongyu Li, Bin Yang, Linlin Ding, Guodong |
| author_facet | Li, Gengliang Chen, Rongyu Li, Bin Yang, Linlin Ding, Guodong |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_15154 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | MedFact-R1: Towards Factual Medical Reasoning via Pseudo-Label Augmentation Li, Gengliang Chen, Rongyu Li, Bin Yang, Linlin Ding, Guodong Computer Vision and Pattern Recognition 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. |
| title | MedFact-R1: Towards Factual Medical Reasoning via Pseudo-Label Augmentation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.15154 |