1 paper
Jaewan Choi, Junyoung Yang, Sangdon Park
Machine unlearning in Large Language Models (LLMs) faces a critical trade-off between erasing target knowledge and preserving general utility. We propose SAUL (Sharpness-Aware Augm…