作者位次 #5/6ACL 2026Poster· Natural Language Processing
H i E dit: Lifelong Model Editing with Hierarchical Reinforcement Learning
Yangfan Wang · Tianyang Sun · Chen Tang · Jie Liu · Wei Cai · Jingchi Jiang
摘要
Lifelong model editing (LME) aims to sequentially rectify outdated or inaccurate knowledge in deployed LLMs while minimizing side effects on unrelated inputs. However, existing approaches typically apply parameter perturbations to a static and dense set of LLM layers for all editing instances. Th...