The Curse of Popularity: Popular Entities have Catastrophic Side Effects when Deleting Knowledge from Language Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866929380254547968 |
|---|---|
| author | Takahashi, Ryosuke Kamoda, Go Heinzerling, Benjamin Sakaguchi, Keisuke Inui, Kentaro |
| author_facet | Takahashi, Ryosuke Kamoda, Go Heinzerling, Benjamin Sakaguchi, Keisuke Inui, Kentaro |
| contents | Language models (LMs) encode world knowledge in their internal parameters through training. However, LMs may learn personal and confidential information from the training data, leading to privacy concerns such as data leakage. Therefore, research on knowledge deletion from LMs is essential. This study focuses on the knowledge stored in LMs and analyzes the relationship between the side effects of knowledge deletion and the entities related to the knowledge. Our findings reveal that deleting knowledge related to popular entities can have catastrophic side effects. Furthermore, this research is the first to analyze knowledge deletion in models trained on synthetic knowledge graphs, indicating a new direction for controlled experiments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_06032 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | The Curse of Popularity: Popular Entities have Catastrophic Side Effects when Deleting Knowledge from Language Models Takahashi, Ryosuke Kamoda, Go Heinzerling, Benjamin Sakaguchi, Keisuke Inui, Kentaro Computation and Language Language models (LMs) encode world knowledge in their internal parameters through training. However, LMs may learn personal and confidential information from the training data, leading to privacy concerns such as data leakage. Therefore, research on knowledge deletion from LMs is essential. This study focuses on the knowledge stored in LMs and analyzes the relationship between the side effects of knowledge deletion and the entities related to the knowledge. Our findings reveal that deleting knowledge related to popular entities can have catastrophic side effects. Furthermore, this research is the first to analyze knowledge deletion in models trained on synthetic knowledge graphs, indicating a new direction for controlled experiments. |
| title | The Curse of Popularity: Popular Entities have Catastrophic Side Effects when Deleting Knowledge from Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2406.06032 |