Non-Gaussianity Reveals Interpretable Directions in Language Models
Finding human-interpretable directions in language-model activations is important for understanding and controlling model behavior. These directions are usually discovered through indirect learning objectives such as sparse reconstruction, but it remains unclear what existing statistical structure distinguishes interpretable directions and whether that structure can be used directly for feature discovery. We find that non-Gaussianity provides such a signature: projection distributions of natural-text residual-stream activations along sparse autoencoder (SAE) decoder directions are substantially more non-Gaussian than those along random directions, with most contexts forming a concentrated...
| Authors | Sida Liu, Feijiang Han, Sarah Wiegreffe et al. |
|---|---|
| Primary category | cs.LG |
| Published | 10 June 2026 |
| Last updated at source | 19 October 2026 |
Popularity, last 4 days
0 now. Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
- sida/ica-lens-paperDataset · cited by
Recent changes
- 20 Oct 2026 Name: ICA Lens: Interpreting Language Models Without Training Another Dictionary → Non-Gaussianity Reveals Interpretable Directions in Language Models
- 20 Oct 2026 Description: Finding interpretable directions in language-model representations is critical for understanding and controlling model behavior. Sparse autoencoders (SAEs) have become the standard tool for this purpose, but using them as the default first → Finding human-interpretable directions in language-model activations is important for understanding and controlling model behavior. These directions are usually discovered through indirect learning objectives such as sparse reconstruction,
Changes detected by Silytics between refreshes of the source.
Models and datasets referencing this paper
- sida/ica-lens-paperDataset
Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 20 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.