KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models
返回详情
VLA / Vision-Language-Action 每日论文卡