Fetching the paper…

Safeguarding Vision-Language Models Against Patched Visual Prompt Injectors · Around