Reproduce “ContactBench 1.0: A Benchmark for Everyday Contact-Rich Robotic 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
Jingyi Xiang, Yixuan Wang, Jeannette Bohg, Wenzhen Yuan
ContactBench benchmarks contact-rich manipulation with eight overlapping skills, 36 reference tasks, printable objects, and simulation models. Compliant PLA springs set resistance independently of fit. For tested door handles, printed forces were 5.01–9.54 N versus 5.45–9.53 N for real handles. Across six tasks, the best policy result was 17/30 (56.7%) on screw tightening; force reproducibility applies only to tested assets, printers, and settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.