MetaKE：面向更优准确性-可编辑性权衡的元学习知识编辑方法

View PDF HTML (experimental)

Abstract:Existing locate-then-edit Knowledge Editing (KE) methods typically decompose editing into two stages: upstream target representation optimization and downstream constrained parameter optimization. The optimization across the two stages is disconnected: upstream applies uniform regularization without observing downstream realization of the planned residual, hindering a refined accuracy-editability trade-off. Since this realization is request-specific and depends on downstream constraints, uniform regularization can over-shrink high-association requests, causing insufficient editing, while it can under-regularize low-association requests, producing over-large planned residuals that reduce downstream editability. To bridge this disconnect, we propose MetaKE (Meta-learning for Knowledge Editing), a new framework that unifies upstream and downstream stages into a bi-level optimization problem. The inner level optimizes parameter updates for the target representation, while the outer level optimizes representation using feedback from downstream constraints, achieving a better semantic accuracy-editability trade-off. To avoid costly multi-layer backpropagation, we introduce a Structural Gradient Proxy to approximate and propagate this feedback. Extensive experiments show that MetaKE outperforms strong baselines, offering a new perspective on KE.

Subjects:	Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as:	arXiv:2603.12677 [cs.CL]
	(or arXiv:2603.12677v4 [cs.CL] for this version)
	https://doi.org/10.48550/arXiv.2603.12677 arXiv-issued DOI via DataCite

Submission history

From: Shuxin Liu [view email]
[v1] Fri, 13 Mar 2026 05:47:00 UTC (371 KB)
[v2] Mon, 16 Mar 2026 01:43:55 UTC (371 KB)
[v3] Thu, 7 May 2026 12:53:41 UTC (2,220 KB)
[v4] Fri, 23 Oct 2026 01:32:57 UTC (2,223 KB)