Related papers: A Comparative Analysis of Sparse Autoencoder and A…
In this paper we propose Structuring AutoEncoders (SAE). SAEs are neural networks which learn a low dimensional representation of data which are additionally enriched with a desired structure in this low dimensional space. While traditional…
Sparse autoencoders (SAEs) have become a standard tool for mechanistic interpretability in autoregressive large language models (LLMs), enabling researchers to extract sparse, human-interpretable features and intervene on model behavior.…
A key challenge in interpretability is to decompose model activations into meaningful features. Sparse autoencoders (SAEs) have emerged as a promising tool for this task. However, a central problem in evaluating the quality of SAEs is the…
It is assumed that sparse autoencoders (SAEs) decompose polysemantic activations into interpretable linear directions, as long as the activations are composed of sparse linear combinations of underlying features. However, we find that if an…
For large language models (LLMs), sparse autoencoders (SAEs) have been shown to decompose intermediate representations that often are not interpretable directly into sparse sums of interpretable features, facilitating better control and…
Understanding the internal machinations of deep Transformer-based NLP models is more crucial than ever as these models see widespread use in various domains that affect the public at large, such as industry, academia, finance, health. While…
Large language models have achieved remarkable capabilities across diverse tasks, yet their internal decision-making processes remain largely opaque, limiting our ability to inspect, control, and systematically improve them. This opacity…
Sparse dictionary learning (and, in particular, sparse autoencoders) attempts to learn a set of human-understandable concepts that can explain variation on an abstract space. A basic limitation of this approach is that it neither exploits…
Sparse Autoencoders (SAEs) have demonstrated significant promise in interpreting the hidden states of language models by decomposing them into interpretable latent directions. However, training and interpreting SAEs at scale remains…
Large Language Model (LLM) deployment requires guiding the LLM to recognize and not answer unsafe prompts while complying with safe prompts. Previous methods for achieving this require adjusting model weights along with other expensive…
Identifying the features learned by neural networks is a core challenge in mechanistic interpretability. Sparse autoencoders (SAEs), which learn a sparse, overcomplete dictionary that reconstructs a network's internal activations, have been…
Sparse autoencoders (SAEs) are used to disentangle the dense, polysemantic internal representations of large language models (LLMs) into interpretable, monosemantic concepts. However, standard $\ell_1$-regularized SAEs suffer from feature…
Large-scale text-to-image diffusion models have become the backbone of modern image editing, yet text prompts alone do not offer adequate control over the editing process. Two properties are especially desirable: disentanglement, where…
Sparse autoencoders (SAEs) have shown promise in extracting interpretable features from complex neural networks. We present one of the first applications of SAEs to dense text embeddings from large language models, demonstrating their…
Sparse autoencoders are a promising new approach for decomposing language model activations for interpretation and control. They have been applied successfully to vision transformer image encoders and to small-scale diffusion models.…
We investigate whether sparse autoencoders (SAEs) can be used to remove knowledge from language models. We use the biology subset of the Weapons of Mass Destruction Proxy dataset and test on the gemma-2b-it and gemma-2-2b-it language…
Sparse Autoencoders (SAEs) that can accurately reconstruct their input (minimizing distortion) by making efficient use of few features (minimizing the rate) often fail to learn monosemantic representations (highly interpretable), limiting…
The linear representation hypothesis states that neural network activations encode high-level concepts as linear mixtures. However, under superposition, this encoding is a projection from a higher-dimensional concept space into a…
Vision foundation models (FMs) achieve state-of-the-art performance in medical imaging. However, they encode information in abstract latent representations that clinicians cannot interrogate or verify. The goal of this study is to…
Sparse autoencoders (SAEs) are increasingly used to extract activation directions for inference-time steering, but their standard sparsity objective treats latent features as independent. This prior can be poorly matched to high-level…