Training-Free Token Pruning via Zeroth-Order Gradient Estimation in Vision-Language Models

VLMRe-assign community