Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.24502 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915826460786688 |
|---|---|
| author | Park, Haewon Kim, Sangwoo Jo, Yohan |
| author_facet | Park, Haewon Kim, Sangwoo Jo, Yohan |
| contents | Knowledge editing aims to efficiently correct factual errors in language models. Widely used locate-then-edit methods update an MLP layer by adjusting its weights to change the mapping between the layer's input vector (key) and output vector (value), thereby editing the model's knowledge. As this update is driven by key and value vectors, obtaining these vectors without careful constraints causes significant model perturbations beyond the targeted edit, a common issue in many prior knowledge editing methods. To address this, we propose Subspace Knowledge Edit (SUIT), which computes key and value vectors only within the subspace of critical features relevant to the edit. Our empirical results on LLaMA3, GPT-J, and Qwen2.5 models show that SUIT dramatically improves knowledge preservation over strong baselines while maintaining high editing performance. These results support the claim that SUIT successfully identifies the critical subspace for the edit. Beyond quantitative gains, our analyses show that SUIT reduces unintended perturbations in hidden states while confining updates to directions that are more effective for editing. Taken together, these findings establish edit-critical subspace identification as a key principle for reliable, low-perturbation knowledge editing. Our code is available at https://github.com/holi-lab/SUIT. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_24502 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SUIT: Knowledge Editing with Subspace-Aware Key-Value Mappings Park, Haewon Kim, Sangwoo Jo, Yohan Computation and Language Knowledge editing aims to efficiently correct factual errors in language models. Widely used locate-then-edit methods update an MLP layer by adjusting its weights to change the mapping between the layer's input vector (key) and output vector (value), thereby editing the model's knowledge. As this update is driven by key and value vectors, obtaining these vectors without careful constraints causes significant model perturbations beyond the targeted edit, a common issue in many prior knowledge editing methods. To address this, we propose Subspace Knowledge Edit (SUIT), which computes key and value vectors only within the subspace of critical features relevant to the edit. Our empirical results on LLaMA3, GPT-J, and Qwen2.5 models show that SUIT dramatically improves knowledge preservation over strong baselines while maintaining high editing performance. These results support the claim that SUIT successfully identifies the critical subspace for the edit. Beyond quantitative gains, our analyses show that SUIT reduces unintended perturbations in hidden states while confining updates to directions that are more effective for editing. Taken together, these findings establish edit-critical subspace identification as a key principle for reliable, low-perturbation knowledge editing. Our code is available at https://github.com/holi-lab/SUIT. |
| title | SUIT: Knowledge Editing with Subspace-Aware Key-Value Mappings |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2509.24502 |