Reproduce “SymAnt: Grounding Long-Horizon Action Anticipation with Adaptable Symbolic Priors” 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

Fnu Aryan, Yaqi Xie, Joseph Campbell, Woojun Kim, Kwonjoon Lee, Hossein Nourkhiz Mahjoub, Ehsan Moradi-Pari, Katia Sycara and 1 more

SymAnt anticipates future actions from brief video observations. It links a temporal scene graph of objects and spatial relations to a knowledge graph of affordances and actions, then uses searched context to iteratively refine an action sequence. On 50 Salads, at 5% observation and 50% prediction, it scores 14.33 MoC versus NeSCA’s 9.38. It does not demonstrate robot assistance.

Summary by alphaXiv.

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