Reproduce “ULTRADISCOVERY: Abductive Exploration in an Interconnected, Epistemically Open Universe” 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
Weihan Li, Tianshi Zheng, Yangqiu Song, Ginny Y. Wong, Simon See
ULTRADISCOVERY tests whether agents can build scientific representations and combine evidence across five domains. Disclosure supplies structural guidance. Disclosure adds an estimated 1.0 grounded finding, while baseline agents retract the taught axiom in 46.1% of Open episodes without constructing its replacement. No system predicts exactly within 200 actions; one aided instance succeeds at 600, while extended Open episodes remain inexact.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.