MoVA: Learning Asymmetric Dual Projections for Modular Long Video-Text Alignment figure
AlphaXiv 中文概览(可滚动查看)