Split, Unlearn, Merge: Leveraging Data Attributes for More Effective Unlearning in LLMs
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_ | 1866914836790640640 |
|---|---|
| author | Kadhe, Swanand Ravindra Ahmed, Farhan Wei, Dennis Baracaldo, Nathalie Padhi, Inkit |
| author_facet | Kadhe, Swanand Ravindra Ahmed, Farhan Wei, Dennis Baracaldo, Nathalie Padhi, Inkit |
| contents | Large language models (LLMs) have shown to pose social and ethical risks such as generating toxic language or facilitating malicious use of hazardous knowledge. Machine unlearning is a promising approach to improve LLM safety by directly removing harmful behaviors and knowledge. In this paper, we propose "SPlit, UNlearn, MerGE" (SPUNGE), a framework that can be used with any unlearning method to amplify its effectiveness. SPUNGE leverages data attributes during unlearning by splitting unlearning data into subsets based on specific attribute values, unlearning each subset separately, and merging the unlearned models. We empirically demonstrate that SPUNGE significantly improves the performance of two recent unlearning methods on state-of-the-art LLMs while maintaining their general capabilities on standard academic benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_11780 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Split, Unlearn, Merge: Leveraging Data Attributes for More Effective Unlearning in LLMs Kadhe, Swanand Ravindra Ahmed, Farhan Wei, Dennis Baracaldo, Nathalie Padhi, Inkit Machine Learning Artificial Intelligence Computation and Language Large language models (LLMs) have shown to pose social and ethical risks such as generating toxic language or facilitating malicious use of hazardous knowledge. Machine unlearning is a promising approach to improve LLM safety by directly removing harmful behaviors and knowledge. In this paper, we propose "SPlit, UNlearn, MerGE" (SPUNGE), a framework that can be used with any unlearning method to amplify its effectiveness. SPUNGE leverages data attributes during unlearning by splitting unlearning data into subsets based on specific attribute values, unlearning each subset separately, and merging the unlearned models. We empirically demonstrate that SPUNGE significantly improves the performance of two recent unlearning methods on state-of-the-art LLMs while maintaining their general capabilities on standard academic benchmarks. |
| title | Split, Unlearn, Merge: Leveraging Data Attributes for More Effective Unlearning in LLMs |
| topic | Machine Learning Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2406.11780 |