Group-SAE: Efficient Training of Sparse Autoencoders for Large Language Models via Layer Groups

TL;DR

Group-SAE introduces a layer grouping strategy for training sparse autoencoders in large language models, significantly reducing training time while maintaining performance and interpretability.

Contribution

The paper proposes Group-SAE, a novel layer grouping method guided by the AMAD metric, to efficiently train SAEs across multiple layers of large language models.

Findings

Significantly accelerates SAE training with minimal quality loss.

Maintains comparable downstream task performance.

Provides a scalable approach for large models.

Abstract

SAEs have recently been employed as a promising unsupervised approach for understanding the representations of layers of Large Language Models (LLMs). However, with the growth in model size and complexity, training SAEs is computationally intensive, as typically one SAE is trained for each model layer. To address such limitation, we propose \textit{Group-SAE}, a novel strategy to train SAEs. Our method considers the similarity of the residual stream representations between contiguous layers to group similar layers and train a single SAE per group. To balance the trade-off between efficiency and performance, we further introduce \textit{AMAD} (Average Maximum Angular Distance), an empirical metric that guides the selection of an optimal number of groups based on representational similarity across layers. Experiments on models from the Pythia family show that our approach significantly accelerates training with minimal impact on reconstruction quality and comparable downstream task performance and interpretability over baseline SAEs trained layer by layer. This method provides an efficient and scalable strategy for training SAEs in modern LLMs.