MM-SafetyBench

anchor

Findings

• IC-021Vision-language adaptation degrades safety in Llama-2-chat-7b even when training data is filtered for safety [eval]
• IC-022Safety layers in Llama-2-chat-7b show substantial divergence during VL adaptation, correlating with safety degradation [eval]
• IC-571Open-source VLMs (LLaVA, MiniGPT-4, InstructBLIP) are substantially more vulnerable to multimodal jailbreak attacks than Gemini-1.5-flash, with BAP attack ASR of 58–62% versus 40–41% [eval]