Papers by Gyubin Choi
Context-Robust Knowledge Editing for Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing knowledge editing methods assess success by considering only edited knowledge without preceding contexts. |
| Approach: | They propose a method to strengthen context robustness by minimizing context-sensitive variance in hidden states of the model. |
| Outcome: | The proposed method improves the success rate in situations where a preceding context is present and preserves the overall capabilities of the model. |