Reproduce “From Human Videos to Bimanual Dexterous Manipulation Data: Confidence-Guided RL Refinement of Video Priors” 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
Yanghong Liu, Yang Chen, Kailang Xu, Bang Du, Feiyang Yang, Jiawei Xia, Zhengtao Yao, Yubiao Zhang and 3 more
PriDex transfers human-video guidance to robot manipulation by using contact regions to select robot-specific grasps and reinforcement learning to refine motion under robot constraints. It weights object trajectories and relative wrist-motion guidance according to reconstruction consistency. In simulation, it averages 57.2% success across four tasks, above adapted baselines, though wrist guidance helps some tasks and hurts others.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.