Latent Knowledge Scalpel: Precise and Massive Knowledge Editing for Large Language Models
Xin Liu, Qiyang Song, Shaowen Xu, Kerou Zhou, Wenbo Jiang, Xiaoqi Jia, Weijuan Zhang, Heqing Huang, Yakai Li · 2025
Read Latent Knowledge Scalpel: Precise and Massive Knowledge Editing for Large Language Models by Xin Liu, Qiyang Song, Shaowen Xu, Kerou Zhou, Wenbo Jiang, Xiaoqi Jia, Weijuan Zhang, Heqing Huang, Yakai Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-03741.txt.
Description
Xin Liu and co-authors edit the facts inside a large language model the way one edits text, in a 2025 paper. Models freeze in wrong or stale facts from training. Older editing tools break when asked to fix thousands of facts at once, or they dull the model's general skill. The team's tests first showed that swapping an entity inside the model's inner representation works much like swapping a word in a prompt.
Their editor, Latent Knowledge Scalpel, steers those inner representations through a small helper network. On two open models it landed ten thousand edits at once while general ability held up. The authors post their code for others to test the claim.
Catalog record
| CREATOR | Xin Liu, Qiyang Song, Shaowen Xu, Kerou Zhou, Wenbo Jiang, Xiaoqi Jia, Weijuan Zhang, Heqing Huang, Yakai Li |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-03741