AI RESEARCH
HiEdit: Lifelong Model Editing with Hierarchical Reinforcement Learning
arXiv CS.CL
•
ArXi:2604.11214v1 Announce Type: new Lifelong model editing (LME) aims to sequentially rectify outdated or inaccurate knowledge in deployed LLMs while minimizing side effects on unrelated inputs.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. This practice is counter-intuitive, as we hypothesize that different pieces of knowledge are d in distinct layers of the model.