Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866918033080975360 |
|---|---|
| author | Fel, Thomas Lubana, Ekdeep Singh Prince, Jacob S. Kowal, Matthew Boutin, Victor Papadimitriou, Isabel Wang, Binxu Wattenberg, Martin Ba, Demba Konkle, Talia |
| author_facet | Fel, Thomas Lubana, Ekdeep Singh Prince, Jacob S. Kowal, Matthew Boutin, Victor Papadimitriou, Isabel Wang, Binxu Wattenberg, Martin Ba, Demba Konkle, Talia |
| contents | Sparse Autoencoders (SAEs) have emerged as a powerful framework for machine learning interpretability, enabling the unsupervised decomposition of model representations into a dictionary of abstract, human-interpretable concepts. However, we reveal a fundamental limitation: existing SAEs exhibit severe instability, as identical models trained on similar datasets can produce sharply different dictionaries, undermining their reliability as an interpretability tool. To address this issue, we draw inspiration from the Archetypal Analysis framework introduced by Cutler & Breiman (1994) and present Archetypal SAEs (A-SAE), wherein dictionary atoms are constrained to the convex hull of data. This geometric anchoring significantly enhances the stability of inferred dictionaries, and their mildly relaxed variants RA-SAEs further match state-of-the-art reconstruction abilities. To rigorously assess dictionary quality learned by SAEs, we introduce two new benchmarks that test (i) plausibility, if dictionaries recover "true" classification directions and (ii) identifiability, if dictionaries disentangle synthetic concept mixtures. Across all evaluations, RA-SAEs consistently yield more structured representations while uncovering novel, semantically meaningful concepts in large-scale vision models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_12892 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models Fel, Thomas Lubana, Ekdeep Singh Prince, Jacob S. Kowal, Matthew Boutin, Victor Papadimitriou, Isabel Wang, Binxu Wattenberg, Martin Ba, Demba Konkle, Talia Computer Vision and Pattern Recognition Sparse Autoencoders (SAEs) have emerged as a powerful framework for machine learning interpretability, enabling the unsupervised decomposition of model representations into a dictionary of abstract, human-interpretable concepts. However, we reveal a fundamental limitation: existing SAEs exhibit severe instability, as identical models trained on similar datasets can produce sharply different dictionaries, undermining their reliability as an interpretability tool. To address this issue, we draw inspiration from the Archetypal Analysis framework introduced by Cutler & Breiman (1994) and present Archetypal SAEs (A-SAE), wherein dictionary atoms are constrained to the convex hull of data. This geometric anchoring significantly enhances the stability of inferred dictionaries, and their mildly relaxed variants RA-SAEs further match state-of-the-art reconstruction abilities. To rigorously assess dictionary quality learned by SAEs, we introduce two new benchmarks that test (i) plausibility, if dictionaries recover "true" classification directions and (ii) identifiability, if dictionaries disentangle synthetic concept mixtures. Across all evaluations, RA-SAEs consistently yield more structured representations while uncovering novel, semantically meaningful concepts in large-scale vision models. |
| title | Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2502.12892 |