AI RESEARCH
Anatomy of Unlearning: The Dual Impact of Fact Salience and Model Fine-Tuning
arXiv CS.CL
•
ArXi:2602.19612v3 Announce Type: replace Machine Unlearning (MU) enables Large Language Models (LLMs) to remove unsafe or outdated information. However, existing work assumes that all facts are equally forgettable and largely ignores whether the forgotten knowledge originates from pre