Improving Adversarial Transferability in MLLMs via Dynamic Vision-Language Alignment Attack

AAML
Re-assign community