AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910915921707008 |
|---|---|
| author | Fang, Junfeng Jiang, Houcheng Wang, Kun Ma, Yunshan Jie, Shi Wang, Xiang He, Xiangnan Chua, Tat-seng |
| author_facet | Fang, Junfeng Jiang, Houcheng Wang, Kun Ma, Yunshan Jie, Shi Wang, Xiang He, Xiangnan Chua, Tat-seng |
| contents | Large language models (LLMs) often exhibit hallucinations due to incorrect or outdated knowledge. Hence, model editing methods have emerged to enable targeted knowledge updates. To achieve this, a prevailing paradigm is the locating-then-editing approach, which first locates influential parameters and then edits them by introducing a perturbation. While effective, current studies have demonstrated that this perturbation inevitably disrupt the originally preserved knowledge within LLMs, especially in sequential editing scenarios. To address this, we introduce AlphaEdit, a novel solution that projects perturbation onto the null space of the preserved knowledge before applying it to the parameters. We theoretically prove that this projection ensures the output of post-edited LLMs remains unchanged when queried about the preserved knowledge, thereby mitigating the issue of disruption. Extensive experiments on various LLMs, including LLaMA3, GPT2-XL, and GPT-J, show that AlphaEdit boosts the performance of most locating-then-editing methods by an average of 36.7% with a single line of additional code for projection solely. Our code is available at: https://github.com/jianghoucheng/AlphaEdit. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_02355 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models Fang, Junfeng Jiang, Houcheng Wang, Kun Ma, Yunshan Jie, Shi Wang, Xiang He, Xiangnan Chua, Tat-seng Computation and Language Artificial Intelligence Large language models (LLMs) often exhibit hallucinations due to incorrect or outdated knowledge. Hence, model editing methods have emerged to enable targeted knowledge updates. To achieve this, a prevailing paradigm is the locating-then-editing approach, which first locates influential parameters and then edits them by introducing a perturbation. While effective, current studies have demonstrated that this perturbation inevitably disrupt the originally preserved knowledge within LLMs, especially in sequential editing scenarios. To address this, we introduce AlphaEdit, a novel solution that projects perturbation onto the null space of the preserved knowledge before applying it to the parameters. We theoretically prove that this projection ensures the output of post-edited LLMs remains unchanged when queried about the preserved knowledge, thereby mitigating the issue of disruption. Extensive experiments on various LLMs, including LLaMA3, GPT2-XL, and GPT-J, show that AlphaEdit boosts the performance of most locating-then-editing methods by an average of 36.7% with a single line of additional code for projection solely. Our code is available at: https://github.com/jianghoucheng/AlphaEdit. |
| title | AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2410.02355 |