Reproduce “CLADE: Category-Level Answering from Deficient Evidence in Retrieval-Augmented Language 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

S M Rayeed, Charles V. Stewart, Yao Ma

The authors study resolvable insufficiency: retrieved text names a broad category but not the exact answer, and a direct guess from it is Clade. Tests found family-only pages reduced abstention than silent pages while lowering precision; replacing an answer with its parent also lowered abstention and raised wrong answers. Ontology routing recovered 26.8% end to end in PrimeKG, with family identification limiting recovery.

Summary by alphaXiv.

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