Event
ARPG: Action Representation Pretraining and Generation
1 reports · 1 sources
Yanjinghao Xu and other authors propose the ARPG (Action Representation Pretraining and Generation) framework, which separates action representation learning from action generation: first, an action representation encoder is pretrained on large-scale human manipulation data; then the encoder is frozen, and a 130M-parameter generative backbone is trained in its representation space to learn task-conditioned action distributions. According to the authors, this design aims to scale up complex dexterous manipulation capabilities; the relevant results currently come from the authors' own account and await independent verification.
Report timeline
- ARPG scales dexterous manipulation via action representation pretraining
ARPG proposes separating action representation learning from action generation: first pretrain an action representation encoder on large-scale human manipulation data, then freeze that encoder and train a 130M-parameter generative backbone in its representation space to learn task-conditioned action distributions.
arXiv cs.RO ·