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

Open in new window