Locate-then-edit for Multi-hop Factual Recall under Knowledge Editing

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhang, Zhuoran, Li, Yongxiang, Kan, Zijian, Cheng, Keyuan, Hu, Lijie, Wang, Di
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913673607380992
author Zhang, Zhuoran
Li, Yongxiang
Kan, Zijian
Cheng, Keyuan
Hu, Lijie
Wang, Di
author_facet Zhang, Zhuoran
Li, Yongxiang
Kan, Zijian
Cheng, Keyuan
Hu, Lijie
Wang, Di
contents The locate-then-edit paradigm has shown significant promise for knowledge editing (KE) in Large Language Models (LLMs). While previous methods perform well on single-hop fact recall tasks, they consistently struggle with multi-hop factual recall tasks involving newly edited knowledge. In this paper, leveraging tools in mechanistic interpretability, we first identify that in multi-hop tasks, LLMs tend to retrieve knowledge with implicit subject information from deeper MLP layers, unlike single-hop tasks, which rely on shallow layers. This distinction explains the poor performance of current methods in multi-hop queries, as they primarily focus on editing shallow layers with single-hop edit prompts, leaving deeper layers unchanged. To address this, we propose IFMET, a novel locate-then-edit KE approach designed to edit both shallow and deep MLP layers. Beyond single-hop editing prompts, IFMET further incorporates multi-hop editing prompts to locate and modify knowledge across different stages of reasoning. Experimental results demonstrate that IFMET significantly improves performance on multi-hop factual recall tasks, overcoming the limitations of previous locate-then-edit methods
format Preprint
id arxiv_https___arxiv_org_abs_2410_06331
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Locate-then-edit for Multi-hop Factual Recall under Knowledge Editing
Zhang, Zhuoran
Li, Yongxiang
Kan, Zijian
Cheng, Keyuan
Hu, Lijie
Wang, Di
Computation and Language
Artificial Intelligence
Machine Learning
The locate-then-edit paradigm has shown significant promise for knowledge editing (KE) in Large Language Models (LLMs). While previous methods perform well on single-hop fact recall tasks, they consistently struggle with multi-hop factual recall tasks involving newly edited knowledge. In this paper, leveraging tools in mechanistic interpretability, we first identify that in multi-hop tasks, LLMs tend to retrieve knowledge with implicit subject information from deeper MLP layers, unlike single-hop tasks, which rely on shallow layers. This distinction explains the poor performance of current methods in multi-hop queries, as they primarily focus on editing shallow layers with single-hop edit prompts, leaving deeper layers unchanged. To address this, we propose IFMET, a novel locate-then-edit KE approach designed to edit both shallow and deep MLP layers. Beyond single-hop editing prompts, IFMET further incorporates multi-hop editing prompts to locate and modify knowledge across different stages of reasoning. Experimental results demonstrate that IFMET significantly improves performance on multi-hop factual recall tasks, overcoming the limitations of previous locate-then-edit methods
title Locate-then-edit for Multi-hop Factual Recall under Knowledge Editing
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2410.06331