Reproduce “Are Frontier VLM Agents Ready to Be Robot Generalists? An Empirical Study with the Embodied Agent Arena” 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

Haojian Huang, Pukun Zhao, Zexi Li, Yehang Zhang, Yangkai Wei, Wenqian Li, Han Yang, Kaiwen Zhou and 2 more

The Embodied Agent Arena evaluates ten models in five domains through fixed-input probes and interactive tasks. It keeps decisions with each model via source-native interfaces, then independently checks whether terminal conditions hold. Astra and Qwen-Max have similar Spatial Pass@1 scores, 69.4% and 68.9%, but Planning Pass@1 scores of 80.3% and 36.9%. On CALVIN, 9 of 100 episodes complete all five goals.

Summary by alphaXiv.

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