DB-KSVD: Scalable Alternating Optimization for Disentangling High-Dimensional Embedding Spaces

TL;DR

This paper introduces DB-KSVD, a scalable dictionary learning algorithm inspired by KSVD, capable of disentangling high-dimensional embeddings in large datasets, and demonstrates its effectiveness on transformer and image models.

Contribution

Proposes DB-KSVD, a scalable adaptation of KSVD for high-dimensional, large-scale dictionary learning, matching SAE performance with a different optimization approach.

Findings

DB-KSVD effectively disentangles text and image embeddings.

It achieves competitive results on SAEBench metrics.

Traditional optimization methods can scale to large, high-dimensional datasets.

Abstract

Dictionary learning has recently emerged as a promising approach for mechanistic interpretability of large transformer models. Disentangling high-dimensional transformer embeddings requires algorithms that scale to high-dimensional data with large sample sizes. Recent work has explored sparse autoencoders (SAEs) for this problem. However, SAEs use a simple linear encoder to solve the sparse encoding subproblem, which is known to be NP-hard. It is therefore interesting to understand whether this approach is sufficient to find good solutions to the dictionary learning problem or if a more sophisticated algorithm could find better solutions. In this work, we propose Double-Batch KSVD (DB-KSVD), a scalable dictionary learning algorithm that adapts the classic KSVD algorithm. DB-KSVD is informed by the rich theoretical foundations of KSVD but scales to datasets with millions of samples and thousands of dimensions. We demonstrate the efficacy of DB-KSVD by disentangling text embeddings of the Gemma-2-2B and Pythia-160M models and evaluating on six metrics from the SAEBench benchmark, where we achieve competitive results when compared to established approaches based on SAEs. We further show similar results when disentangling image embeddings obtained from the DINOv2-S and DINOv2-B models, solidifying our findings. By matching SAE performance with an entirely different optimization approach, our results suggest that (i) SAEs do find strong solutions to the dictionary learning problem and (ii) traditional optimization approaches can be scaled to the required problem sizes, offering a promising avenue for further research. We make an implementation of DB-KSVD available at https://github.com/romeov/ksvd.jl.