Interpretability-Guided Layer Selection over Subspace Projection: SAEs as Stethoscopes, Not Scalpels, for Raw Task Vector Model Editing
Li LeiIncept LabsHouston, TXMadalina CiobanuIncept LabsHouston, TX, Qingqing MaoIncept Labs, Houston, TXTitan Holdings, San Francisco, CA [email protected] Ritankar DasIncept Labs, Houston, TXTitan Holdings, San Francisco, CA · 2026
Read Interpretability-Guided Layer Selection over Subspace Projection: SAEs as Stethoscopes, Not Scalpels, for Raw Task Vector Model Editing by Li LeiIncept LabsHouston, TXMadalina CiobanuIncept LabsHouston, TX, Qingqing MaoIncept Labs, Houston, TXTitan Holdings, San Francisco, CA [email protected] Ritankar DasIncept Labs, Houston, TXTitan Holdings, San Francisco, CA (2026) 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-2605-28649.txt.
Description
Li Lei and three colleagues test sparse-autoencoder-guided editing for math reasoning on Gemma-3-4B-IT. Projecting task vectors onto autoencoder feature subspaces discards about 97 percent of the modification energy, giving no significant gains across seven math subjects. The failure traces to geometric mismatch between activation-space autoencoder directions and weight-space task vectors. Using autoencoders only to pick layers for unfiltered task vectors lifts Number Theory accuracy from about 30 to about 39 percent, with five subjects improved and none degraded.
Catalog record
| CREATOR | Li LeiIncept LabsHouston, TXMadalina CiobanuIncept LabsHouston, TX, Qingqing MaoIncept Labs, Houston, TXTitan Holdings, San Francisco, CA [email protected] Ritankar DasIncept Labs, Houston, TXTitan Holdings, San Francisco, CA |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46238 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.28649