arXiv cs.CL· Diego Miguel Lozano, Erik Derner, Juan Antonio P\'erez-Ortiz, Nuria Oliver·· 12 小时前AI 评分42
LLM 安全中的极性盲区：否定、反义词替换与对抗性包装的系统性研究
Polarity Blindness in LLM Safety: A Systematic Study of Negation, Antonym Substitution, and Adversarial Framing
AI 导读
研究提出“极性盲区”这一结构性漏洞：LLM 在语法否定、反义词替换等改变极性的语言变换后，无法识别有害意图。基于 260 条提示词、10 个 SOTA 模型和英语/西班牙语/捷克语 3 种语言（按 0-5 危害量表评估），多数模型受影响，最脆弱模型平均危害分达 4.12。该效应在不同否定体系的语言中均存在，说明这是语义鲁棒性的普遍局限而非特定语法问题。
正文 · 原文
Abstract:Negation remains a persistent challenge for language models across NLP tasks and modalities. We show that this weakness can undermine LLM safety guardrails through a structural vulnerability we term polarity blindness: the failure to recognize harmful intent after polarity-altering linguistic transformations. We study two forms of polarity inversion: grammatical negation and antonym replacement, and use synonym replacement as a polarity-preserving control. We further test whether an adversarial textual wrapper amplifies these failures. Using a 260-prompt benchmark across 10 state-of-the-art LLMs and 3 languages (English, Spanish, Czech), stratified by harmfulness potential and evaluated on a validated 0-5 harm scale, we find that polarity blindness affects most evaluated models, with mean harm scores reaching 4.12 for the most vulnerable model. The effect persists across typologically differing negation systems, suggesting that polarity blindness reflects a broader limitation in semantic robustness rather than a language- or grammar-specific failure.
| Comments: | Code and data: this https URL |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.27908 [cs.CL] |
| (or arXiv:2610.27908v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.27908 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Diego Miguel Lozano [view email]
[v1]
Tue, 20 Oct 2026 09:22:50 UTC (2,318 KB)
来源：arXiv cs.CL · arxiv.org