English(EN) SpecGuard: Inference-Time Backdoor Detection For Free

SpecGuard 使用推测解码技术免费检测 LLM 后门

作者 PulseAugur 编辑部 · [2 个来源] · 2026-09-10 16:51

研究人员开发了 SpecGuard，一种在推理过程中检测大型语言模型后门的新颖方法。该技术重新利用了通常用于加速模型生成的推测解码过程，以识别恶意行为，而无需额外的计算成本。SpecGuard 利用草稿模型预测与目标模型验证过程之间的差异来检测触发的后门，即使是那些设计隐蔽的后门。 AI

影响该方法通过在推理过程中检测隐藏的后门，提供了一种免费增强 LLM 安全性的方式。

排序理由该集群描述了一篇详细介绍 LLM 后门检测新方法的论文。

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Research; 该集群描述了一篇详细介绍 LLM 后门检测新方法的论文。
Source corroboration: 2 independent sources; Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
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: 27 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

arXiv cs.CL TIER_1 English(EN) · Rui Wen, Ahmed Salem, Andrew Paverd, Mark Russinovich, Zheng Li · 2026-09-11 04:00

SpecGuard：免费的推理时后门检测

arXiv:2609.11799v1 Announce Type: cross Abstract: Large language models are often fine-tuned, shared, or downloaded from third parties, so a deployed model may carry a hidden backdoor that behaves normally on benign inputs but switches to attacker-controlled behavior when a secre…
Hugging Face Daily Papers TIER_1 English(EN) · 2026-09-10 16:51

SpecGuard：免费的推理时后门检测

Large language models are often fine-tuned, shared, or downloaded from third parties, so a deployed model may carry a hidden backdoor that behaves normally on benign inputs but switches to attacker-controlled behavior when a secret trigger appears. While backdoors can be audited …