Reproduce “Gym-Anything for Robotics: Scaling physical simulations with digital agents” 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
Pranjal Aggarwal, Sean Welleck
Gym-Anything for Robotics uses coding agents to build robotics simulations and tasks, then separate auditors test their evidence and probe for defects. The system selects 120 action groups and builds 25 simulations, forming 200 RoboWorld environments. Claude Opus 5.5 succeeds on 0% of 25 test tasks, while reaching 32% on an easier training subset; these are simulated tasks, not real-robot results.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.