Fetching the paper…

VLATTACK: Multimodal Adversarial Attacks on Vision-Language Tasks via Pre-trained Models · Around