AI RESEARCH
MetaKE: Meta-learning Aligned Knowledge Editing via Bi-level Optimization
arXiv CS.AI
•
ArXi:2603.12677v1 Announce Type: cross Knowledge editing (KE) aims to precisely rectify specific knowledge in Large Language Models (LLMs) without disrupting general capabilities. State-of-the-art methods suffer from an open-loop control mismatch. We identify a critical "Semantic-Execution Disconnect": the semantic target is derived independently without feedback from the downstream's feasible region. This misalignment often causes valid semantic targets to fall within the prohibited space, resulting in gradient truncation and editing failure.