Reproduce “Steertuning Policies from a Few Human Videos” 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
Billy Yan, Isaac Tu, Kelly Lee, Blake Chang, Kenny Ma, Kanad Patel, Dhawal Kabra, Neer Patel and 3 more
Steertuning trains a diffusion policy on annotated egocentric human video, then uses five task demonstrations to adapt it and steer action generation at deployment. The demonstrations help the policy learn a task and guide its predictions. Across 12 task-environment combinations, the authors report 71% average success, while noting that visual-only annotations omit contact information and limit dexterity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.