An Image Is Worth 1000 Lies: Transferability of Adversarial Images across Prompts on Vision-Language Models.

Haochen Luo, Jindong Gu, Fengyuan Liu, Philip Torr

Browse the full ICLR paper archive.