Reproduce “Training Reasoning Models in Simulation to Natively Use Real Robots” 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
Jiaming Tang, Zhuoyang Zhang, Junxian Guo, Fangzhou Zhao, Yilong Zhao, Zekai Li, Xingyang Li, Yuyang Zhao and 3 more
The authors train a robot-use model with reinforcement learning on simulated tasks, using task-success rewards and a shared end-effector pose interface. The model observes camera images and robot state, then proposes each move. It reaches 74.4% on LIBERO and succeeds in 18 of 30 real YAM trials, versus none for its base model; evidence is limited to tested tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.