Reproduce “DexFormer 2.0: Towards a Cross-Embodiment Foundation Policy for Dexterous Grasping” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Lixin Xu, Ke Zhang, Xinyu Liu, Jie Tan, Senming Tan, Jiaju Yin, Yating Feng, Xiaoyi Lin and 2 more
DexFormer 2.0 shares a dexterous-grasping policy across arm–hand pairs by mapping robots into 29 canonical joint slots, encoding joint morphology, and using 16 recent interaction steps. On ten seen primitive objects, it achieves 51.18% ± 1.72 stable success on six held-out pairings, whose arm and hand types appeared in training; this does not establish transfer to wholly unseen hardware.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.