← 研踪 PaperTrailACL Anthology · ACL 2026:acl:2026.acl-long.240 · 记录 #1
Reasoning Structure Matters for Safety Alignment of Reasoning Models
来源日期（仅到月份） 2026-07
来源记录日期 2026-07
雷达观测 2026-11-01 22:04
摘要
Large reasoning models (LRMs) achieve strong performance on complex reasoning tasks but often generate harmful responses to malicious user queries. This paper investigates the underlying cause of these safety risks and shows that the issue lies in the reasoning structure itself. Based on this insight, we claim that effective safety alignment can be achieved by altering the reasoning structure. We propose AltTrain, a simple yet effective post-training method that explicitly alters the reasoning structure of LRMs. AltTrain is both practical and generalizable, requiring no complex reinforcement learning (RL) training or reward design—only supervised fine-tuning (SFT) with a lightweight 1K training examples. Experiments across LRM backbones and model sizes demon strate strong safety alignment, along with robust generalization across reasoning, QA, summarization, and multilingual setting.
论文发现出处
这些页面提供论文线索，不代表同行评审或质量认证。
观测到的版本
平台记录号代表本站观测到的元数据变化，不是作者修订号。来源日期仅到月份或年份时，按该月或该年首日存档。
记录 #1 · 2026-07 · 最新观测版本
Reasoning Structure Matters for Safety Alignment of Reasoning Models
Large reasoning models (LRMs) achieve strong performance on complex reasoning tasks but often generate harmful responses to malicious user queries. This paper investigates the underlying cause of these safety risks and shows that the issue lies in the reasoning structure itself. Based on this insight, we claim that effective safety alignment can be achieved by altering the reasoning structure. We propose AltTrain, a simple yet effective post-training method that explicitly alters the reasoning structure of LRMs. AltTrain is both practical and generalizable, requiring no complex reinforcement learning (RL) training or reward design—only supervised fine-tuning (SFT) with a lightweight 1K training examples. Experiments across LRM backbones and model sizes demon strate strong safety alignment, along with robust generalization across reasoning, QA, summarization, and multilingual setting.
查看此版本解读 →查看此版本原文 ↗