O-Edit: Orthogonal Subspace Editing for Language Model Sequential Editing
Fuente:
arXiv
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910651383808000 |
|---|---|
| author | Cai, Yuchen Cao, Ding |
| author_facet | Cai, Yuchen Cao, Ding |
| contents | Large language models (LLMs) acquire knowledge during pre-training, but over time, this knowledge may become incorrect or outdated, necessitating updates after training. Knowledge editing techniques address this issue without the need for costly re-training. However, most existing methods are designed for single edits, and as the number of edits increases, they often cause a decline in the model's overall performance, posing significant challenges for sequential editing. To overcome this, we propose Orthogonal Subspace Editing, O-Edit. This algorithm orthogonalizes the direction of each knowledge update, minimizing interference between successive updates and reducing the impact of new updates on unrelated knowledge. Our approach does not require replaying previously edited data and processes each edit knowledge on time. It can perform thousands of edits on mainstream LLMs, achieving an average performance improvement that is 4.2 times better than existing methods while effectively preserving the model's performance on downstream tasks, all with minimal additional parameter overhead. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_11469 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | O-Edit: Orthogonal Subspace Editing for Language Model Sequential Editing Cai, Yuchen Cao, Ding Computation and Language Large language models (LLMs) acquire knowledge during pre-training, but over time, this knowledge may become incorrect or outdated, necessitating updates after training. Knowledge editing techniques address this issue without the need for costly re-training. However, most existing methods are designed for single edits, and as the number of edits increases, they often cause a decline in the model's overall performance, posing significant challenges for sequential editing. To overcome this, we propose Orthogonal Subspace Editing, O-Edit. This algorithm orthogonalizes the direction of each knowledge update, minimizing interference between successive updates and reducing the impact of new updates on unrelated knowledge. Our approach does not require replaying previously edited data and processes each edit knowledge on time. It can perform thousands of edits on mainstream LLMs, achieving an average performance improvement that is 4.2 times better than existing methods while effectively preserving the model's performance on downstream tasks, all with minimal additional parameter overhead. |
| title | O-Edit: Orthogonal Subspace Editing for Language Model Sequential Editing |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2410.11469 |