Improving Dictionary Learning with Gated Sparse Autoencoders

arXiv:2404.16014v1 Announce Type: new
Abstract: Recent work has found that sparse autoencoders (SAEs) are an effective technique for unsupervised discovery of interpretable features in language models' (LMs) activations, by finding sparse, linear reconstructions of LM activations. We introduce the Gated Sparse Autoencoder (Gated SAE), which achieves a Pareto improvement over training with prevailing methods. In SAEs, the L1 penalty used to encourage sparsity introduces many undesirable biases, such as shrinkage -- systematic underestimation of feature activations. The key insight of …

abstract arxiv autoencoder autoencoders cs.ai cs.lg dictionary discovery features found improvement improving language language models linear lms pareto training type unsupervised work