Reproduce “Mycelium: A Meta-Study of Hallucination Detection Across Provenance, Typology, and Granularities” 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

Julia Belikova, Elisei Rykov

Mycelium examines how benchmark design shapes scores for hallucination detectors across provenance, typology, and granularity. The authors compare annotation targets and metrics, and propose Span Coverage F1 for nested span boundaries. In Mu-SHROOM leave-one-annotator-out analysis, its macro score rises from 0.410 to 0.750 against majority and union references, while exact span F1 falls; this measures reference sensitivity, not detector gains.

Summary by alphaXiv.

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