Fetching the paper…

An Image Is Worth 1000 Lies: Adversarial Transferability across Prompts on Vision-Language Models · Around