Skip to content
AI Atlas
PaperActive

When Auditors Fabricate: Batch-Size Degradation and Confident Hallucination in LLM Detection of Planted Document Contamination

arxiv.org/abs/2609.09696

Updated 52 min ago · first seen 11 Sept 2026

paper_01M294GMKKKZHV3QQNE03652RB

Published
11 Sept 2026
T1 · 52 min ago
arXiv
2609.09696
T1 · 52 min ago
Category
cs.CL
T1 · 52 min ago

Abstract

Large language models are increasingly proposed as automated auditors of document quality, yet their reliability as detectors of planted errors is poorly characterised. We construct a contaminated corpus of 150 academic papers spanning supply chain management and medical research, injecting 450 known contaminants of three types: typographical corruption, semantic reversal, and absurd out-of-context insertion. We then evaluate Google Gemini 3.0 Pro's ability to recover a 180-contaminant answer-key subset across 60 documents under three prompting regimes of increasing scale: single document, small batch, and large batch. Detection holds at small scale and then collapses: 50% recovery on single documents, 60% on small batches, and 2.8% on large batches. The failure mode at scale is not abstention but fabrication. Rather than reporting incomplete processing, the model produced confident findings including invented contaminants of its own, absurdities such as "telepathic squirrel" and "quantum-powered toaster" that mimic the style of the planted material but do not appear in any document. Detection also varies by contamination type: absurd insertions were recovered at 75% in completed evaluations, while semantic reversals and typographical corruptions were each recovered at only 50%. The corruptions most likely to occur in the wild, plausible ones, are the ones most often missed. We conclude that LLM document auditing degrades not gracefully but deceptively, and outline the harness such systems require: bounded batch sizes, direct content injection, and mechanical verification of every reported finding against source text.

Authors 4

Karan Parekh, Sanjana Pendyala Ravinder, Sana Mhapsekar, Medina Maloku

Specification

Official page

Source:arXiv (Atom API + RSS)T1observed 52 min agohigh

Arxiv announce type
cross

Source:arXiv (Atom API + RSS)T1observed 52 min agohigh

arXiv id
2609.09696

Source:arXiv (Atom API + RSS)T1observed 52 min agohigh

Categories
cs.CL, cs.AI

Source:arXiv (Atom API + RSS)T1observed 52 min agohigh

PDF

Source:arXiv (Atom API + RSS)T1observed 52 min agohigh

Primary category
cs.CL

Source:arXiv (Atom API + RSS)T1observed 52 min agohigh

Published
11 Sept 2026

Source:arXiv (Atom API + RSS)T1observed 52 min agohigh

Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →

Provenance

Attributed facts

9

Source tiers

T19

Freshest observation

52 min ago

Conflicts

None