MedVLThinker: Simple Baselines for Multimodal Medical Reasoning

TL;DR

MedVLThinker introduces an open, reproducible framework for medical multimodal reasoning, demonstrating that text-only data training under reinforcement learning yields superior performance and setting new benchmarks in medical question answering.

Contribution

The paper provides a simple, strong baseline recipe for medical multimodal reasoning models, including data curation, training paradigms, and open-source resources, advancing reproducibility and performance.

Findings

RLVR outperforms SFT in experiments.

Training on text-only data improves performance more than multimodal data.

The 7B model achieves state-of-the-art results on medical VQA benchmarks.

Abstract

Large Reasoning Models (LRMs) have introduced a new paradigm in AI by enabling models to ``think before responding" via chain-of-thought reasoning. However, the absence of open and reproducible recipes for building reasoning-centric medical LMMs hinders community-wide research, analysis, and comparison. In this paper, we present MedVLThinker, a suite of simple yet strong baselines. Our fully open recipe consists of: (1) systematic data curation for both text-only and image-text medical data, filtered according to varying levels of reasoning difficulty, and (2) two training paradigms: Supervised Fine-Tuning (SFT) on distilled reasoning traces and Reinforcement Learning with Verifiable Rewards (RLVR) based on final answer correctness. Across extensive experiments on the Qwen2.5-VL model family (3B, 7B) and six medical QA benchmarks, we find that RLVR consistently and significantly outperforms SFT. Additionally, under the RLVR framework, a key, counter-intuitive finding is that training on our curated text-only reasoning data provides a more substantial performance boost than training on multimodal image-text data. Our best open 7B model, trained using the RLVR recipe on text-only data, establishes a new state-of-the-art on existing public VQA benchmarks, surpassing all previous open-source medical LMMs. Furthermore, scaling our model to 32B achieves performance on par with the proprietary GPT-4o. We release all curated data, models, and code to provide the community with a strong, open foundation for future research in multimodal medical reasoning.