Teach Old SAEs New Domain Tricks with Boosting
This enables researchers to selectively enhance SAE interpretability for specific domains, addressing a bottleneck in targeted mechanistic interpretability of LLMs, though it is incremental as it builds on existing SAE methods.
The paper tackles the problem of sparse autoencoders (SAEs) failing to capture domain-specific features in large language models by introducing a residual learning approach that trains a secondary SAE to model reconstruction errors on domain-specific texts. The result shows significant improvements in LLM cross-entropy and explained variance metrics across multiple specialized domains while maintaining general task performance.
Sparse Autoencoders have emerged as powerful tools for interpreting the internal representations of Large Language Models, yet they often fail to capture domain-specific features not prevalent in their training corpora. This paper introduces a residual learning approach that addresses this feature blindness without requiring complete retraining. We propose training a secondary SAE specifically to model the reconstruction error of a pretrained SAE on domain-specific texts, effectively capturing features missed by the primary model. By summing the outputs of both models during inference, we demonstrate significant improvements in both LLM cross-entropy and explained variance metrics across multiple specialized domains. Our experiments show that this method efficiently incorporates new domain knowledge into existing SAEs while maintaining their performance on general tasks. This approach enables researchers to selectively enhance SAE interpretability for specific domains of interest, opening new possibilities for targeted mechanistic interpretability of LLMs.