Detecting and Correcting Reference Hallucinations: 3-13% of LLM Citation URLs Never Existed

arXiv·high signal

Systematic measurement of citation URL reliability across 10 models and 4 deep research agents using DRBench (53,090 URLs) and ExpertQA (168,021 URLs across 32 academic fields). Finds 3-13% of citation URLs are hallucinated — no record in the Wayback Machine and likely never existed — while 5-18% are no longer accessible. First study to quantify the scope of reference hallucination across commercial LLMs and the emerging class of deep research agents.