新的ALE方法可高效估算微调LLM中的罕见故障率

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-28 04:00

研究人员开发了一种名为Amplify, Locate, Estimate (ALE)的新方法，用于高效测量微调大型语言模型中的罕见故障率。该技术利用token对数概率的变化来放大不良行为的发生，从而能够以更低的成本和更快的速度估算低至$10^{-5}$的故障概率。ALE在41个评估目标上均表现出很高的准确性，其中位数估计值与参考值之比接近1，与直接测量方法相比，显著减少了所需的测试轮次。 AI

影响能够对微调LLM进行更高效、更准确的安全评估，这对于在敏感应用中部署模型至关重要。

排序理由该集群包含一篇详细介绍LLM新评估方法的论文。[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: 12 / 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新评估方法的论文。[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: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Patrick Reichherzer, Jonathan Evertz, Thorsten Holz · 2026-10-28 04:00

对微调LLM稀有故障率的放大、定位和估计 (ALE)

arXiv:2610.35472v1 Announce Type: new Abstract: Safety fine-tuning can make undesirable model behaviors sufficiently rare that measuring their frequency by direct sampling becomes prohibitively expensive. We introduce Amplify, Locate, Estimate (ALE), a method for estimating such …

本文如何被排名

报道来源 [1]

对微调LLM稀有故障率的放大、定位和估计 (ALE)

相关实体

相关话题