MixTok: Adaptive Visual Tokenization with Global 1D Tokens and Sparse 2D Refinement figure
AlphaXiv 中文概览(可滚动查看)