Making Every Step Effective: Jailbreaking Large Vision-Language Models Through Hierarchical KV Equalization

AAML
VLM
Re-assign community