Reproduce “Unifying Deep Predicate Invention with Pre-trained Foundation Models” 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
Qianwei Wang, Bowen Li, Zhanpeng Luo, Yifan Xu, Alexander Gray, Tom Silver, Sebastian Scherer, Katia Sycara and 1 more
UniPred uses an LLM to propose predicates and skill effects, then tests and refines them with learning from demonstrations. Its iterative feedback loop averages 94.2% held-out success across four simulated domains, compared with 25.0% for a single-completion variant. On a real robot, it demonstrates table cleaning; results remain bounded by deterministic STRIPS-style controller assumptions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.