Reproduce “Coding Agents with Harness for Safe Robot Control” 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

Bingxin Xu, Yuzhang Shang, Zhen Dong, Emilio Ferrara

SafeHarness adds obstacle checks to coding-agent robot manipulation during travel and contact. It verifies proposed routes before execution and selects contact approaches based on nearby obstacles. On SafeLIBERO, GPT-6-Astra averages 81.2% task success and 91.9% collision-avoidance rates. Against a matched no-harness setting, it improves both measures; these rates track obstacle displacement, not every physical contact.

Summary by alphaXiv.

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