To address the challenges posed by the high computational load of Vision Transformers, various lightweight methods such as pruning and quantization have been proposed. Among them, token pruning aims ...