Aligning Language Models with Real-time Knowledge Editing

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Tang, Chenming, Yang, Yutong, Wang, Kexue, Wu, Yunfang
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911617567948800
author Tang, Chenming
Yang, Yutong
Wang, Kexue
Wu, Yunfang
author_facet Tang, Chenming
Yang, Yutong
Wang, Kexue
Wu, Yunfang
contents Knowledge editing aims to modify outdated knowledge in language models efficiently while retaining their original capabilities. Mainstream datasets for knowledge editing are predominantly static and fail to keep in pace with the evolving real-world knowledge. In this work, we introduce CRAFT, an ever-evolving real-world dataset for knowledge editing. It evaluates models on temporal locality, common-sense locality, composite portability and alias portability, providing a comprehensive and challenging evaluation for knowledge editing, on which previous methods hardly achieve balanced performance. Towards flexible real-time knowledge editing, we propose KEDAS, a novel paradigm of knowledge editing alignment featuring diverse edit augmentation and self-adaptive post-alignment inference, exhibiting significant performance gain on both CRAFT and traditional datasets compared to previous methods. We hope this work may serve as a catalyst for shifting the focus of knowledge editing from static update to dynamic evolution.
format Preprint
id arxiv_https___arxiv_org_abs_2508_01302
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Aligning Language Models with Real-time Knowledge Editing
Tang, Chenming
Yang, Yutong
Wang, Kexue
Wu, Yunfang
Computation and Language
Computational Engineering, Finance, and Science
Knowledge editing aims to modify outdated knowledge in language models efficiently while retaining their original capabilities. Mainstream datasets for knowledge editing are predominantly static and fail to keep in pace with the evolving real-world knowledge. In this work, we introduce CRAFT, an ever-evolving real-world dataset for knowledge editing. It evaluates models on temporal locality, common-sense locality, composite portability and alias portability, providing a comprehensive and challenging evaluation for knowledge editing, on which previous methods hardly achieve balanced performance. Towards flexible real-time knowledge editing, we propose KEDAS, a novel paradigm of knowledge editing alignment featuring diverse edit augmentation and self-adaptive post-alignment inference, exhibiting significant performance gain on both CRAFT and traditional datasets compared to previous methods. We hope this work may serve as a catalyst for shifting the focus of knowledge editing from static update to dynamic evolution.
title Aligning Language Models with Real-time Knowledge Editing
topic Computation and Language
Computational Engineering, Finance, and Science
url https://arxiv.org/abs/2508.01302