Our new preprint on Safety of Vision Language Models is out. (Update Jan 23, 2025: Accepted to ICLR 2025!)