04:00
2026-09-10
arxiv.org
large-language-models
When Auditors Fabricate: Batch-Size Degradation and Confident Hallucination in LLM Detection of Planted Document Contamination
Google Gemini 3.0 Pro recovered only 2.8% of a 180-contaminant answer key when auditing large batches of documents, down from 50% on single documents and 60% on small batches, according to an arXiv pa…