Fast Enough to Act: Spatio-Temporal Visual Token Merging for Low-Latency Robotic VLMs and VLAs figure
AlphaXiv 中文概览(可滚动查看)