Preprint
Oct 2026
Scaling Dexterous Manipulation via Action Representation Pretraining
The proposed ARPG (Action Representation Pretraining and Generation), which explicitly separates action representation learning from action generation, achieves state-of-the-art performance on the DexJoCo benchmark and improves the average success rate on bimanual dexterous manipulation.
Yanjinghao Xu, Jie-Xin Zhang, Huan Zhao et al.
· 0 citations