Saved in:
Bibliographic Details
Main Authors: Su, Yuhao, Choudhuri, Anwesa, Gao, Zhongpai, Planche, Benjamin, Nguyen, Van Nguyen, Zheng, Meng, Shen, Yuhan, Innanje, Arun, Chen, Terrence, Elhamifar, Ehsan, Wu, Ziyan
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.06581
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913014595190784
author Su, Yuhao
Choudhuri, Anwesa
Gao, Zhongpai
Planche, Benjamin
Nguyen, Van Nguyen
Zheng, Meng
Shen, Yuhan
Innanje, Arun
Chen, Terrence
Elhamifar, Ehsan
Wu, Ziyan
author_facet Su, Yuhao
Choudhuri, Anwesa
Gao, Zhongpai
Planche, Benjamin
Nguyen, Van Nguyen
Zheng, Meng
Shen, Yuhan
Innanje, Arun
Chen, Terrence
Elhamifar, Ehsan
Wu, Ziyan
contents Large vision-language models struggle with medical video understanding, where spatial precision, temporal reasoning, and clinical semantics are critical. To address this, we first introduce \textbf{MedVidBench}, a large-scale benchmark of 531,850 video-instruction pairs across 8 medical sources spanning video, segment, and frame-level tasks, curated through a rigorous quality assurance pipeline with expert-guided prompting and dual-model validation. While supervised fine-tuning on MedVidBench yields noticeable gains, standard Reinforcement Learning (RL) fails due to imbalanced reward scales across datasets, which destabilizes optimization and leads to training collapse. To overcome this, we introduce \textbf{MedGRPO}, a novel RL framework for balanced multi-dataset training with two key innovations: (1) \emph{cross-dataset reward normalization} that maps each dataset's median performance to a common reward value, ensuring fair optimization regardless of difficulty, and (2) a \emph{medical LLM judge} that evaluates caption quality on five clinical dimensions through comparative similarity scoring. Supervised fine-tuning Qwen2.5-VL-7B on MedVidBench outperforms GPT-4.1 and Gemini-2.5-Flash across all tasks, while MedGRPO further improves the SFT baseline on grounding and captioning. Our work establishes a foundational benchmark and training methodology for advancing medical video understanding with VLMs. Our project website is available at: https://uii-america.github.io/MedGRPO/.
format Preprint
id arxiv_https___arxiv_org_abs_2512_06581
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MedGRPO: Multi-Task Reinforcement Learning for Heterogeneous Medical Video Understanding
Su, Yuhao
Choudhuri, Anwesa
Gao, Zhongpai
Planche, Benjamin
Nguyen, Van Nguyen
Zheng, Meng
Shen, Yuhan
Innanje, Arun
Chen, Terrence
Elhamifar, Ehsan
Wu, Ziyan
Computer Vision and Pattern Recognition
Large vision-language models struggle with medical video understanding, where spatial precision, temporal reasoning, and clinical semantics are critical. To address this, we first introduce \textbf{MedVidBench}, a large-scale benchmark of 531,850 video-instruction pairs across 8 medical sources spanning video, segment, and frame-level tasks, curated through a rigorous quality assurance pipeline with expert-guided prompting and dual-model validation. While supervised fine-tuning on MedVidBench yields noticeable gains, standard Reinforcement Learning (RL) fails due to imbalanced reward scales across datasets, which destabilizes optimization and leads to training collapse. To overcome this, we introduce \textbf{MedGRPO}, a novel RL framework for balanced multi-dataset training with two key innovations: (1) \emph{cross-dataset reward normalization} that maps each dataset's median performance to a common reward value, ensuring fair optimization regardless of difficulty, and (2) a \emph{medical LLM judge} that evaluates caption quality on five clinical dimensions through comparative similarity scoring. Supervised fine-tuning Qwen2.5-VL-7B on MedVidBench outperforms GPT-4.1 and Gemini-2.5-Flash across all tasks, while MedGRPO further improves the SFT baseline on grounding and captioning. Our work establishes a foundational benchmark and training methodology for advancing medical video understanding with VLMs. Our project website is available at: https://uii-america.github.io/MedGRPO/.
title MedGRPO: Multi-Task Reinforcement Learning for Heterogeneous Medical Video Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.06581