一作ACL 2026Findings· Findings of ACL
When Models Outthink Their Safety: Unveiling and Mitigating Self-Jailbreak in Large Reasoning Models
Yingzhi Mao · Chunkang Zhang · Junxiang Wang · Xinyan Guan · Boxi Cao · Yaojie Lu · Hongyu Lin · Xianpei Han · Le Sun
摘要
Large Reasoning Models (LRMs) achieve strong performance on complex multi-step reasoning, yet they still exhibit severe safety failures such as harmful content generation. Existing methods often apply coarse-grained constraints over the entire reasoning trajectories, which can undermine reasoning...