Diagnosing Hidden Instabilities in Model Editing via Uncertainty Quantification
Zihan Gu · TianYi Zhang · Xinyan Zhang · Zhiyuan Wang · Han Zhang · Yuhao Wei · Jiacheng Lu · Tianyi Ma · Xingsheng Zhang · Hua Zhang · Yue Hu
摘要
Model editing provides a promising mechanism for updating large language models (LLMs) without expensive retraining. Existing approaches, particularly locate-and-edit methods based on least-squares optimization, aim to introduce targeted knowledge changes while preserving pre-trained behavior. In...