SA-Attack: Improving Adversarial Transferability of Vision-Language Pre-training Models via Self-Augmentation

AAML
VLM
Re-assign community