Reproduce “Words2Contact: Identifying Support Contacts from Verbal Instructions Using Foundation Models” 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

Dionis Totsila, Quentin Rouxel, Jean-Baptiste Mouret, Serena Ivaldi

Words2Contact lets an operator specify and revise a humanoid robot’s support-contact point in language, then confirms it before a whole-body controller executes the motion. In a 78-example benchmark, its selected model combination achieved median success rates of 0.71 for absolute and 0.50 for relative locations, versus 0.12 and 0.17 for random points. These are pixel-selection results, not evidence that surfaces are safe.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.