Definition
Plain language
An automated tool that reads a paper's reference list and checks whether each cited work actually exists.
As stated in the literature
A funnel-shaped citation-verification pipeline that extracts references from a PDF, matches them against six bibliographic catalogs, and escalates only the unresolved residue to a web-search-equipped LLM for evidence-gathering, flagging fabrications and author-identity corruption at roughly four cents per paper.
Why it matters: It catches fabricated or corrupted citations cheaply and at scale, helping keep invented references out of the published record.
For example, RefChecker can scan a paper's bibliography, confirm most entries against reference databases, and send only the doubtful few to a web search for closer inspection.
Heard on the show
“The tool is called RefChecker, and it's a funnel.”Episode 201 — One in Four NeurIPS Papers Cites a Reference That Doesn't Exist