DocMEdit: Towards Document-Level Model Editing

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zeng, Li, Liu, Zeming, Feng, Chong, Huang, Heyan, Guo, Yuhang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916759177527296
author Zeng, Li
Liu, Zeming
Feng, Chong
Huang, Heyan
Guo, Yuhang
author_facet Zeng, Li
Liu, Zeming
Feng, Chong
Huang, Heyan
Guo, Yuhang
contents Model editing aims to correct errors and outdated knowledge in the Large language models (LLMs) with minimal cost. Prior research has proposed a variety of datasets to assess the effectiveness of these model editing methods. However, most existing datasets only require models to output short phrases or sentences, overlooks the widespread existence of document-level tasks in the real world, raising doubts about their practical usability. Aimed at addressing this limitation and promoting the application of model editing in real-world scenarios, we propose the task of document-level model editing. To tackle such challenges and enhance model capabilities in practical settings, we introduce \benchmarkname, a dataset focused on document-level model editing, characterized by document-level inputs and outputs, extrapolative, and multiple facts within a single edit. We propose a series of evaluation metrics and experiments. The results show that the difficulties in document-level model editing pose challenges for existing model editing methods.
format Preprint
id arxiv_https___arxiv_org_abs_2505_19572
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DocMEdit: Towards Document-Level Model Editing
Zeng, Li
Liu, Zeming
Feng, Chong
Huang, Heyan
Guo, Yuhang
Computation and Language
Artificial Intelligence
Model editing aims to correct errors and outdated knowledge in the Large language models (LLMs) with minimal cost. Prior research has proposed a variety of datasets to assess the effectiveness of these model editing methods. However, most existing datasets only require models to output short phrases or sentences, overlooks the widespread existence of document-level tasks in the real world, raising doubts about their practical usability. Aimed at addressing this limitation and promoting the application of model editing in real-world scenarios, we propose the task of document-level model editing. To tackle such challenges and enhance model capabilities in practical settings, we introduce \benchmarkname, a dataset focused on document-level model editing, characterized by document-level inputs and outputs, extrapolative, and multiple facts within a single edit. We propose a series of evaluation metrics and experiments. The results show that the difficulties in document-level model editing pose challenges for existing model editing methods.
title DocMEdit: Towards Document-Level Model Editing
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.19572