Back to Search Start Over

A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders

Authors :
Chanin, David
Wilken-Smith, James
Dulka, Tomáš
Bhatnagar, Hardik
Bloom, Joseph
Publication Year :
2024

Abstract

Sparse Autoencoders (SAEs) have emerged as a promising approach to decompose the activations of Large Language Models (LLMs) into human-interpretable latents. In this paper, we pose two questions. First, to what extent do SAEs extract monosemantic and interpretable latents? Second, to what extent does varying the sparsity or the size of the SAE affect monosemanticity / interpretability? By investigating these questions in the context of a simple first-letter identification task where we have complete access to ground truth labels for all tokens in the vocabulary, we are able to provide more detail than prior investigations. Critically, we identify a problematic form of feature-splitting we call feature absorption where seemingly monosemantic latents fail to fire in cases where they clearly should. Our investigation suggests that varying SAE size or sparsity is insufficient to solve this issue, and that there are deeper conceptual issues in need of resolution.

Details

Database :
arXiv
Publication Type :
Report
Accession number :
edsarx.2409.14507
Document Type :
Working Paper