English(EN) Expected Harm: Rethinking Safety Evaluation of (Mis)Aligned LLMs

新指标“预期危害”挑战大型语言模型安全评估

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-02 05:00

一篇新研究论文提出了一个名为“预期危害”的指标，用于重新评估大型语言模型的安全评估方式。作者认为，当前方法过分强调有害输出的严重性，同时低估了现实世界行为者实现这些输出的可能性。他们的研究表明，模型通常会拒绝低概率威胁，但仍然容易受到高概率威胁的影响，这种校准不当可能源于监督微调数据。 AI

影响提出了一个新的评估大型语言模型安全性的框架，可能带来更强大、更可靠的AI系统。

排序理由介绍大型语言模型安全评估新指标的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

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: 1 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 介绍大型语言模型安全评估新指标的研究论文。[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: 1 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

arXiv cs.CL TIER_1 English(EN) · Yen-Shan Chen, Zhi Rui Tam, Cheng-Kuang Wu, Yun-Nung Chen · 2026-11-02 05:00

Expected Harm: Rethinking Safety Evaluation of (Mis)Aligned LLMs

arXiv:2602.01600v2 Announce Type: replace-cross Abstract: Current evaluations of LLM safety predominantly rely on severity-based taxonomies to assess the harmfulness of models' responses to malicious queries. We argue that this formulation requires re-examination as it implicitly…