Empirical Analysis of Large Vision-Language Models against Goal Hijacking via Visual Prompt Injection

MLLM
Re-assign community