Query-Relevant Images Jailbreak Large Multi-Modal Models
Query-Relevant Images Jailbreak Large Multi-Modal Models
Xin Liu,Yichen Zhu,2 저자,Yu Qiao
2023 · DOI: 10.48550/arXiv.2311.17600
arXiv.org · 51회 인용
TLDR
A novel visual prompt attack that exploits query-relevant images to jailbreak the open-source LMMs and shows LLMs can be easily at-tacked by this approach, even if the employed Large Language Models are safely aligned.