DriveTok: 3D Driving Scene Tokenization for Unified Multi-View Reconstruction and Understanding

返回详情 VLA / Vision-Language-Action 每日论文卡