Our new preprint, How Does Vision-Language Adaptation Impact the Safety of Vision Language Models? is out.