Related Knowledge Perturbation Matters: Rethinking Multiple Pieces of Knowledge Editing in Same-Subject
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866929709311328256 |
|---|---|
| author | Duan, Zenghao Duan, Wenbin Yin, Zhiyi Shen, Yinghan Jing, Shaoling Zhang, Jie Shen, Huawei Cheng, Xueqi |
| author_facet | Duan, Zenghao Duan, Wenbin Yin, Zhiyi Shen, Yinghan Jing, Shaoling Zhang, Jie Shen, Huawei Cheng, Xueqi |
| contents | Knowledge editing has become a promising approach for efficiently and precisely updating knowledge embedded in large language models (LLMs). In this work, we focus on Same-Subject Editing, which involves modifying multiple attributes of a single entity to ensure comprehensive and consistent updates to entity-centric knowledge. Through preliminary observation, we identify a significant challenge: Current state-of-the-art editing methods struggle when tasked with editing multiple related knowledge pieces for the same subject. To address the lack of relevant editing data for identical subjects in traditional benchmarks, we introduce the $\text{S}^2\text{RKE}$(Same-Subject Related Knowledge Editing) benchmark. Our extensive experiments reveal that only mainstream locate-then-edit methods, such as ROME and MEMIT, exhibit "related knowledge perturbation," where subsequent edits interfere with earlier ones. Further analysis reveals that these methods over-rely on subject information, neglecting other critical factors, resulting in reduced editing effectiveness. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_06868 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Related Knowledge Perturbation Matters: Rethinking Multiple Pieces of Knowledge Editing in Same-Subject Duan, Zenghao Duan, Wenbin Yin, Zhiyi Shen, Yinghan Jing, Shaoling Zhang, Jie Shen, Huawei Cheng, Xueqi Computation and Language Artificial Intelligence Knowledge editing has become a promising approach for efficiently and precisely updating knowledge embedded in large language models (LLMs). In this work, we focus on Same-Subject Editing, which involves modifying multiple attributes of a single entity to ensure comprehensive and consistent updates to entity-centric knowledge. Through preliminary observation, we identify a significant challenge: Current state-of-the-art editing methods struggle when tasked with editing multiple related knowledge pieces for the same subject. To address the lack of relevant editing data for identical subjects in traditional benchmarks, we introduce the $\text{S}^2\text{RKE}$(Same-Subject Related Knowledge Editing) benchmark. Our extensive experiments reveal that only mainstream locate-then-edit methods, such as ROME and MEMIT, exhibit "related knowledge perturbation," where subsequent edits interfere with earlier ones. Further analysis reveals that these methods over-rely on subject information, neglecting other critical factors, resulting in reduced editing effectiveness. |
| title | Related Knowledge Perturbation Matters: Rethinking Multiple Pieces of Knowledge Editing in Same-Subject |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2502.06868 |