Reproduce “PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation” 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
Kun Song, Yiming Wang, Yilin Chen, Tianyi Ding, Jiaxin Tian, Tianqi Gong, Daolin Ma, Jia Pan
PEARS adapts a frozen robot policy by combining episode-level, vision-language-model-guided updates to force bounds with reinforcement-learning steering of the policy’s input noise and a contact-force controller. On Bottle Cap Twisting, it reached an 87.7% final test success rate and the 80% online-success threshold at episode 22, versus episode 47 for the strongest baseline; this threshold comparison measures episodes, not elapsed time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.