OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features
This addresses representation issues in interpretable AI for researchers, though it is incremental as it builds on existing SAE methods.
The paper tackled the problems of feature absorption and composition in sparse autoencoders (SAEs) by introducing OrtSAE, which enforces orthogonality between features, resulting in 9% more distinct features, 65% reduction in absorption, 15% reduction in composition, and 6% improvement in spurious correlation removal.
Sparse autoencoders (SAEs) are a technique for sparse decomposition of neural network activations into human-interpretable features. However, current SAEs suffer from feature absorption, where specialized features capture instances of general features creating representation holes, and feature composition, where independent features merge into composite representations. In this work, we introduce Orthogonal SAE (OrtSAE), a novel approach aimed to mitigate these issues by enforcing orthogonality between the learned features. By implementing a new training procedure that penalizes high pairwise cosine similarity between SAE features, OrtSAE promotes the development of disentangled features while scaling linearly with the SAE size, avoiding significant computational overhead. We train OrtSAE across different models and layers and compare it with other methods. We find that OrtSAE discovers 9% more distinct features, reduces feature absorption (by 65%) and composition (by 15%), improves performance on spurious correlation removal (+6%), and achieves on-par performance for other downstream tasks compared to traditional SAEs.