Author name: Gouki Minegishi, Hiroki Furuta, Takeshi Kojima, Yusuke Iwasawa, Yutaka Matsuo

arXiv:2605.00842v1 Announce Type: new
Abstract: Emergent misalignment, where fine-tuning on narrow, non-harmful tasks induces harmful behaviors, poses a key challenge for AI safety in LLMs. Despite growing empirical evidence, its underlying mechanism …