Reproduce “OmniGuide: Universal Spatial Guidance Fields for Enhancing Generalist Robot Policies” 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

Yunzhou Song, Long Le, Yong-Hyun Park, Jie Wang, Junyao Shi, Lingjie Liu, Jiatao Gu, Eric Eaton and 2 more

OmniGuide steers a pretrained generalist robot policy by adding external spatial, semantic, or demonstration-based guidance during action generation. It maps predicted actions into shared 3D workspace and uses guidance gradients to attract or repel trajectories while retaining the policy’s behavior. On a static-barrier task, it reaches 90% success, versus 0% for π₀.₅ and 70% for a specialized baseline. Soft guidance gives no strict guarantee.

Summary by alphaXiv.

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