Preregistration v2 — Diagnostic provenance: do LLM fact-checkers use a stated source reliability on knowledge-graph evidence? (P0 provaudit, Option B)
Preregistered analysis plan, frozen before any non-mock generation of this arm. Separate from, and complementary to, the original P0 preregistration (concept DOI 10.5281/zenodo.21346538). Within-item, label-only audit: claim and evidence bytes identical, only the PROV-O provenance label (human-curated vs unverified LLM extraction) varies; the class error rates are stated in the prompt. Files: plan, corpus v3 manifest (sha256), frozen prompt and analysis code.
Authors
- Oussama Chahed (ORCID: https://orcid.org/0009-0009-8062-4377)
Institutions
- Université Paris Sciences et Lettres (FR)
- Université Paris Dauphine-PSL (FR)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-17
- DOI
- https://doi.org/10.5281/zenodo.24311511
- Primary Topic
- Misinformation and Its Impacts
- Type
- preprint