作者位次 #5/8ACL 2026Poster· Natural Language Processing
From Narrow Unlearning to Emergent Misalignment in LLM s
Erum Mushtaq · Anil Ramakrishna · Satyapriya Krishna · Sattvik Sahai · Prasoon Goyal · Kai-Wei Chang · Tao Zhang · Rahul Gupta
摘要
Recent work has shown that fine-tuning on insecure code data can trigger an emergent misalignment (EMA) phenomenon, where models generate malicious responses even to prompts unrelated to the original insecure code-writing task. Such cross-domain generalization of harmful behavior underscores the...