World2Act: Latent Action Post-Training via Skill-Compositional World Models
返回详情
VLA / Vision-Language-Action 每日论文卡