Keys to Robust Edits: from Theoretical Insights to Practical Advances

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yan, Jianhao, Wang, Futing, Luo, Yun, Li, Yafu, Zhang, Yue
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912386158428160
author Yan, Jianhao
Wang, Futing
Luo, Yun
Li, Yafu
Zhang, Yue
author_facet Yan, Jianhao
Wang, Futing
Luo, Yun
Li, Yafu
Zhang, Yue
contents Large language models (LLMs) struggle with maintaining accurate knowledge due to conflicting/outdated parametric memories. While locate-and-edit methods address this, their reliance on models' internal representations leads to robustness failures in long-context reasoning and paraphrased queries. We identify a fundamental limitation of locate-and-edit methods: existing semantic keys (for memory localization) cannot simultaneously satisfy robustness (context-invariant activation) and specificity (precise knowledge discrimination). Through theoretical error-bound analysis, we establish formal criteria for effective editing. Our solution introduces \textit{Robust Edit Pathway (REP)}, a plug-and-play module that: (1) disentangles editing keys from native model representations; (2) dynamically adjusts keys via contrastive learning to achieve robustness-specificity balance. Extensive experiments across various editing methods (ROME/MEMIT/R-ROME/EMMET), existing LLMs (LLaMA2, QWen, Mistral), and datasets (CounterFact, ZsRE) show that REP improves success rate over robustness tests by up-to 66.4\% while maintaining the success rate unaffected. Our code can be found at https://github.com/ElliottYan/RobustKeyEdit .
format Preprint
id arxiv_https___arxiv_org_abs_2410_09338
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Keys to Robust Edits: from Theoretical Insights to Practical Advances
Yan, Jianhao
Wang, Futing
Luo, Yun
Li, Yafu
Zhang, Yue
Computation and Language
Large language models (LLMs) struggle with maintaining accurate knowledge due to conflicting/outdated parametric memories. While locate-and-edit methods address this, their reliance on models' internal representations leads to robustness failures in long-context reasoning and paraphrased queries. We identify a fundamental limitation of locate-and-edit methods: existing semantic keys (for memory localization) cannot simultaneously satisfy robustness (context-invariant activation) and specificity (precise knowledge discrimination). Through theoretical error-bound analysis, we establish formal criteria for effective editing. Our solution introduces \textit{Robust Edit Pathway (REP)}, a plug-and-play module that: (1) disentangles editing keys from native model representations; (2) dynamically adjusts keys via contrastive learning to achieve robustness-specificity balance. Extensive experiments across various editing methods (ROME/MEMIT/R-ROME/EMMET), existing LLMs (LLaMA2, QWen, Mistral), and datasets (CounterFact, ZsRE) show that REP improves success rate over robustness tests by up-to 66.4\% while maintaining the success rate unaffected. Our code can be found at https://github.com/ElliottYan/RobustKeyEdit .
title Keys to Robust Edits: from Theoretical Insights to Practical Advances
topic Computation and Language
url https://arxiv.org/abs/2410.09338