Seeing No Evil: Blinding Large Vision-Language Models to Safety Instructions via Adversarial Attention Hijacking.

Jingru Li, Wei Ren, Tianqing Zhu

Browse the full ACL paper archive.