Reproduce “LLM4Robot-Bench: How Far Can GPT-6 Astra Go in Robotic Manipulation? A Hierarchical, Paired Sim-Real Benchmark” 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

Haibao Yu, Jiaqi Meng, Zijian Zhang, Bowen Jing, Lei Yang, Yue Hu, Chenchen Ge, Yiming Hou and 11 more

LLM4Robot-Bench groups household manipulation tasks by physical interaction demands and evaluates matched tasks in simulation and reality. Its levels progress from useful contact and alignment to force regulation and changing contacts. For GPT-6 Astra, mean success falls from 80.0% to 47.6% to 0.0% in reality, and from 66.7% to 47.6% to 0.0% in simulation; these results are specific to Astra.

Summary by alphaXiv.

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