Reproduce “Grounded Action Model: 3D Grounding as a Foundation for Robotics” 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
Gehao Zhang, Weikai Huang, Shailesh Shailesh, Yiyan Peng, Jiafei Duan, Ranjay Krishna
GAM gives a robot policy an object-centered observation by using a language instruction, screen point, or box with a frozen 3D grounding model to obtain geometry and visual features. It combines these with robot state to predict actions. On YAM, GAM achieved 17/20 under visual shift versus π₀.₅’s 4/20, after both reached 19/20 in-distribution; differing target interfaces limit causal attribution.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.