Score: 0

Evaluating Sparse Autoencoders: From Shallow Design to Matching Pursuit

Published: June 5, 2025 | arXiv ID: 2506.05239v1

By: Valérie Costa , Thomas Fel , Ekdeep Singh Lubana and more

Potential Business Impact:

Finds hidden patterns in handwritten numbers.

Business Areas:
Semantic Search Internet Services

Sparse autoencoders (SAEs) have recently become central tools for interpretability, leveraging dictionary learning principles to extract sparse, interpretable features from neural representations whose underlying structure is typically unknown. This paper evaluates SAEs in a controlled setting using MNIST, which reveals that current shallow architectures implicitly rely on a quasi-orthogonality assumption that limits the ability to extract correlated features. To move beyond this, we introduce a multi-iteration SAE by unrolling Matching Pursuit (MP-SAE), enabling the residual-guided extraction of correlated features that arise in hierarchical settings such as handwritten digit generation while guaranteeing monotonic improvement of the reconstruction as more atoms are selected.

Country of Origin
🇺🇸 United States

Page Count
12 pages

Category
Computer Science:
Machine Learning (CS)