新的TRACER方法训练LLM以防御间接提示注入攻击

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

研究人员开发了TRACER，一种新颖的对抗性训练方法，旨在保护大型语言模型免受间接提示注入攻击。该方法专门解决了攻击嵌入在检索内容或工具输出中比直接消息更易成功这一漏洞。TRACER采用通道感知对抗性训练和奖励塑造机制，使受攻击的推理与干净证据保持一致，显著降低了攻击成功率，同时在干净任务上保持高准确性。该方法显示出有希望的结果，特别是在使用Qwen2.5-3B-Instruct模型的GSM8K基准测试上。 AI

影响增强了LLM抵御复杂提示注入技术的安全性，可能提高了在使用检索内容的应用中的可靠性。

排序理由该集群包含一篇详细介绍LLM安全新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 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: 16 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍LLM安全新方法的论文。[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.LG TIER_1 English(EN) · Kshitij Mishra, Salem Lahlou, Nils Lukas · 2026-10-15 04:00

TRACER：一种用于对抗间接提示注入的轨迹正则化训练方法

arXiv:2610.19133v1 Announce Type: new Abstract: Indirect prompt injection uses untrusted retrieved content, tool outputs, or agent messages to override an LLM's intended behavior. We show that defense robustness is channel-conditioned: the same attack succeeds 18% of the time as …

本文如何被排名

报道来源 [1]

TRACER：一种用于对抗间接提示注入的轨迹正则化训练方法

相关实体

相关话题