Enhancing the Safety of Medical Vision-Language Models by Synthetic Demonstrations

arXiv:2506.09067v2 Announce Type: replace-cross Abstract: Generative medical vision-language models~(Med-VLMs) are primarily designed to generate complex textual information~(e.g., diagnostic reports) from multimodal inputs including vision modality~(e.g., medical images) and language modality~(e.g., clinical queries). However, their security vulnerabilities remain underexplored. Med-VLMs should be capable of rejecting harmful queries, such as extit{Provide detailed instructions for using this CT scan for insurance fraud}. At the same time, addressing security concerns introduces the risk of over-defense, where safety-enhancing mechanisms may degrade general performance, causing Med-VLMs to reject benign clinical queries. In this paper, we propose a novel inference-time defense strategy to mitigate harmful queries, enabling defense against visual and textual jailbreak attacks. Using diverse medical imaging datasets collected from nine modalities, we demonstrate that our defense strategy based on synthetic clinical demonstrations enhances model safety without significantly compromising performance. Additionally, we find that increasing the demonstration budget alleviates the over-defense issue. We then introduce a mixed demonstration strategy as a trade-off solution for balancing security and performance under few-shot demonstration budget constraints.

Related

• Better Eyes, Better Thoughts: Why Vision Chain-of-Thought Fails in Medicine
• Temporal Inversion for Learning Interval Change in Chest X-Rays
• Leave My Images Alone: Preventing Multi-Modal Large Language Models from Analyzing Images via Visual Prompt Injection
• Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs

Source: arXiv cs.AI | 2026-04-13