arXiv cs.CL· Jinhu Fu, Yan Bai, Yihang Lou, Li Sun, Jiahong Wu, Xiangxiang Chu, Sen Su·· 15 小时前AI 评分26
DynEdit：用动态熵引导的大语言模型长序列知识编辑
DynEdit: Dynamic Entropy-Guided Sequential Editing for Large Language Models
AI 导读
DynEdit 提出用 token 级熵引导长序列知识编辑，对高熵难编辑区域做非均匀滑动以提升覆盖，并在每个编辑窗口内对信息密集 token 做熵引导的 token 级优化。在三个 benchmark 和两个骨干模型上，DynEdit 一致优于竞争基线，在困难的复述查询上最高提升 +14.0 BLEU，多个知识领域语义分数接近饱和。该工作已被 NeurIPS 2026 主会接收为 poster。
正文 · 原文
Abstract:Knowledge editing aims to update outdated or incorrect knowledge in large language models without full retraining. Existing parameter-editing methods remain limited for long-sequence knowledge: single-point editors provide insufficient coverage for extended targets, while fixed chunk-based editors distribute editing effort uniformly across rigid predefined windows, failing to prioritize difficult, information-dense tokens. In this work, we show that token entropy is closely associated with both editing difficulty and semantic importance: high-entropy regions are harder to edit and often contain key factual content. Motivated by this observation, we propose DynEdit, a dynamic framework for long-sequence knowledge editing. DynEdit uses token-level entropy to guide non-uniform sliding, increasing the coverage of difficult high-entropy regions, and further applies entropy-guided token-level optimization to strengthen updates on information-dense tokens within each editing window. Experiments on three benchmarks and two backbone models show that DynEdit consistently outperforms competitive baselines across diverse long-form editing settings, achieving gains of up to +14.0 BLEU points on challenging paraphrase queries and near-saturated semantic scores on several diverse knowledge domains.
| Comments: | Accepted by main conference of NeurIPS 2026 (poster) |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.34768 [cs.CL] |
| (or arXiv:2610.34768v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.34768 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jinhu Fu [view email]
[v1]
Mon, 26 Oct 2026 03:13:26 UTC (1,365 KB)
来源：arXiv cs.CL · arxiv.org