Auditing Frontier Vision-Language Models for Trustworthy Medical VQA: Grounding Failures, Format Collapse, and Domain Adaptation

TL;DR

This paper evaluates the trustworthiness of frontier vision-language models in medical visual question answering, highlighting grounding failures, format issues, and the potential for domain adaptation to improve performance.

Contribution

It provides a comprehensive audit of recent models' grounding and accuracy in medical VQA, identifying perception as a key bottleneck and demonstrating domain adaptation benefits.

Findings

Models poorly localize anatomical and pathological targets.

Self-grounding pipelines degrade VQA accuracy due to perception failures.

Supervised fine-tuning improves VQA recall significantly.

Abstract

Deploying vision-language models (VLMs) in clinical settings demands auditable behavior under realistic failure conditions, yet the failure landscape of frontier VLMs on specialized medical inputs is poorly characterized. We audit five recent frontier and grounding-aware VLMs (Gemini~2.5~Pro, GPT-5, o3, GLM-4.5V, Qwen~2.5~VL) on Medical VQA along two trust-relevant axes. Perception: all models localize anatomical and pathological targets poorly -- the best model reaches only 0.23 mean IoU and 19.1% Acc@0.5 -- and exhibit clinically dangerous laterality confusion. Pipeline integration: a self-grounding pipeline, where the same model localizes then answers, degrades VQA accuracy for every model -- driven by both inaccurate localization and format-compliance failures under the two-step prompt (parse failure rises to 70%--99% for Gemini and GPT-5 on VQA-RAD). Replacing predicted boxes with ground-truth annotations recovers and improves VQA accuracy, consistent with the failure residing in the perception module rather than in the decomposition itself. These observational findings identify grounding quality as a primary trustworthiness bottleneck in our SLAKE bounding-box setting. As a complementary fine-tuning follow-up, supervised fine-tuning of Qwen~2.5~VL on combined Med-VQA training data attains the highest reported SLAKE open-ended recall (85.5%) among comparable methods, suggesting that the VQA-level gap is tractable with domain adaptation; whether this also closes the perception/trustworthiness bottleneck is left to future work.