Improving Sparse Autoencoder with Dynamic Attention

Recently, sparse autoencoders (SAEs) have emerged as a promising technique for interpreting activations in foundation models by disentangling features into a sparse set of concepts. However, identifying the optimal level of sparsity for each neuron remains challenging in practice: excessive sparsity might lead to poor reconstruction, whereas insufficient sparsity harms interpretability. While existing activation functions such as ReLU and TopK provide certain sparsity guarantees, they typically require additional sparsity regularization or cherry-picked hyperparameters. We show in this paper that adaptive sparse attention mechanisms using sparsemax can bridge this trade-off, due to their ability to determine the number of concepts in a datadependent manner. Specifically, we first explore a new class of SAEs based on the cross-attention architecture with the latent features as queries and the learnable dictionary as the key and value matrices. To encourage sparse pattern learning, we employ a sparsemax-based attention strategy that automatically infers a sparse set of concepts according to the complexity of each neuron, resulting in a more flexible and efficient activation function. Through comprehensive evaluation and visualization, we show that our approach successfully achieves lower reconstruction loss while producing high-quality concepts. Moreover, the sparsity level automatically determined by our approach can serve as tuning guidance to improve existing SAEs.