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

大型语言模型文档审计规模化退化，捏造错误

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-10 04:00

一项新研究表明，像Google Gemini 3.0 Pro这样的大型语言模型在规模化检测种植文档污染时，会表现出显著的性能下降和幻觉。虽然在单文档级别有效，但当处理大批量文档时，模型识别错误的能力会急剧下降，导致捏造虚假污染物。研究强调，与荒谬的插入相比，大型语言模型在检测拼写错误和语义反转等合理错误方面能力较弱，这表明需要增强基于大型语言模型的审计系统的验证机制和有界批量处理。 AI

影响突出了大型语言模型在自动化审计任务中可靠性的关键限制，并暗示了在需要高精度应用中存在的潜在风险。

排序理由学术论文，详细介绍研究结果。 [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 12 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，详细介绍研究结果。 [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Karan Parekh, Sanjana Pendyala Ravinder, Sana Mhapsekar, Medina Maloku · 2026-09-10 04:00

当审计员捏造：LLM检测植入文档污染中的批次大小退化与自信幻觉

arXiv:2609.09696v1 Announce Type: new 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 spann…

本文如何被排名

报道来源 [1]

当审计员捏造：LLM检测植入文档污染中的批次大小退化与自信幻觉

相关实体

相关话题