SAVeS: Steering Safety Judgments in Vision-Language Models via Semantic Cues

Vision-language models (VLMs) are increasingly deployed in real-world and embodied settings where safety decisions depend on visual context. However, it remains unclear which visual evidence drives these judgments. We study whether multimodal safety behavior in VLMs can be steered by simple semantic cues. We introduce a semantic steering framework that applies controlled textual, visual, and cognitive interventions without changing the underlying scene content. To evaluate these effects, we propose SAVeS, a benchmark for situational safety under semantic cues, together with an evaluation protocol that separates behavioral refusal, grounded safety reasoning, and false refusals. Experiments across multiple VLMs and an additional state-of-the-art benchmark show that safety decisions are highly sensitive to semantic cues, indicating reliance on learned visual-linguistic associations rather than grounded visual understanding. We further demonstrate that automated steering pipelines can exploit these mechanisms, highlighting a potential vulnerability in multimodal safety systems.

Key Contributions

• Introduces SAVeS benchmark for evaluating situational safety under semantic cues with protocol separating behavioral refusal, grounded reasoning, and false refusals
• Demonstrates VLM safety decisions are highly sensitive to semantic cues (textual, visual, cognitive) without changing underlying scene content
• Shows automated steering pipelines can exploit learned visual-linguistic associations to bypass multimodal safety systems

🛡️ Threat Analysis

Details

Similar Papers