English(EN) Negation Neglect: When models fail to learn negations in training

新研究揭示大型语言模型在否定训练中会学习错误信息

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-09 05:00

一项新的研究论文介绍了一种名为“否定忽略”的现象，即大型语言模型在接受明确声明某项主张为假的文档训练时，会悖谬地学会相信该主张为真。这种效应在包括 Qwen3.5-397B-A17B、Kimi K2.5 和 GPT-4.1 在内的多个模型中都有观察到，即使否定被清晰地呈现。研究人员认为，这表明模型存在一种将主张表示为真的归纳偏见，这可能对 AI 安全产生影响，因为模型在接受标记为恶意的训练内容时可能会采纳恶意行为。 AI

影响模型在接受否定或标记内容训练时可能采纳错误信念或恶意行为，对 AI 安全构成风险。

排序理由研究论文详细介绍了一种大型语言模型的新型故障模式。[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: 14 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 研究论文详细介绍了一种大型语言模型的新型故障模式。[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) · Harry Mayne, Lev McKinney, Jan Dubi\'nski, Adam Karvonen, James Chua, Owain Evans · 2026-11-09 05:00

否定遗忘：模型在训练中未能学会否定

arXiv:2605.13829v2 Announce Type: replace Abstract: We introduce Negation Neglect, where finetuning LLMs on documents that flag a claim as false makes them believe the claim is true. For example, models are finetuned on documents that convey "Ed Sheeran won the 100m gold at the 2…

本文如何被排名

报道来源 [1]

否定遗忘：模型在训练中未能学会否定

相关实体

相关话题