English(EN) What the Model Cannot See: Covert and Provable Text Invisibility in LLMs

新技术使文本对大型语言模型不可见，带来安全风险

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-15 04:00

研究人员开发了一种使文本对大型语言模型不可见的方法，有效地创建了一个“跨度否决”机制，导致模型忽略特定内容。这种不可见性可用于多种目的，包括作为供应链后门或用于审计AI系统。虽然这种攻击的简单版本可以通过线性探测来检测，但更复杂的版本可以加固攻击，使其难以检测，从而在探测可见性与路由可见性之间进行权衡。 AI

影响这项研究突显了大型语言模型潜在的漏洞，影响了AI安全和审计实践。

排序理由详细介绍大型语言模型新技术的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 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: 15 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍大型语言模型新技术的学术论文。[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.AI TIER_1 English(EN) · Mark Russinovich · 2026-10-15 04:00

模型无法看到的内容：LLM 中隐蔽且可证明的文本不可见性

arXiv:2610.15883v1 Announce Type: cross Abstract: Text wrapped in a trigger can be made behaviorally invisible to a language model: omitted from summaries and quotations, ignored in question answering, and left unflagged by downstream filtering. We study this selective blindness …

本文如何被排名

报道来源 [1]

模型无法看到的内容：LLM 中隐蔽且可证明的文本不可见性

相关实体

相关话题