English
Related papers

Related papers: Features that Make a Difference: Leveraging Gradie…

200 papers

Mechanistic interpretability of large language models (LLMs) aims to uncover the internal processes of information propagation and reasoning. Sparse autoencoders (SAEs) have demonstrated promise in this domain by extracting interpretable…

Machine Learning · Computer Science 2025-05-26 Wei Shi , Sihang Li , Tao Liang , Mingyang Wan , Guojun Ma , Xiang Wang , Xiangnan He

Latent steering exploits internal representations of Large Language Models (LLMs) to guide generation, yet interventions on dense states can entangle distinct semantic features. In this paper, we investigate attention query activations as a…

Machine Learning · Computer Science 2026-05-25 Sumanta Bhattacharyya , Pedram Rooshenas

Understanding the multilingual mechanisms of large language models (LLMs) provides insight into how they process different languages, yet this remains challenging. Existing studies often focus on individual neurons, but their polysemantic…

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…

Machine Learning · Computer Science 2024-11-05 Eoin Farrell , Yeu-Tong Lau , Arthur Conmy

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…

Machine Learning · Computer Science 2019-08-20 Marco Rudolph , Bastian Wandt , Bodo Rosenhahn

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…

Computation and Language · Computer Science 2025-06-03 Mark Muchane , Sean Richardson , Kiho Park , Victor Veitch

Sparse autoencoders (SAEs) are a mechanistic interpretability technique that have been used to provide insight into learned concepts within large protein language models. Here, we employ TopK and Ordered SAEs to investigate autoregressive…

Recent work on sparse autoencoders (SAEs) has shown promise in extracting interpretable features from neural networks and addressing challenges with polysemantic neurons caused by superposition. In this paper, we apply SAEs to the early…

Machine Learning · Computer Science 2024-09-10 Liv Gorton

Sparse Autoencoder (SAE) features have become essential tools for mechanistic interpretability research. SAE features are typically characterized by examining their activating examples, which are often "monosemantic" and align with human…

Artificial Intelligence · Computer Science 2025-09-30 Claire Tian , Katherine Tian , Nathan Hu

Sparse Autoencoders (SAEs) are widely employed for mechanistic interpretability and model steering. Within this context, steering is by design performed by means of decoding altered SAE intermediate representations. This procedure…

Machine Learning · Computer Science 2025-12-08 Antonio Bărbălau , Cristian Daniel Păduraru , Teodor Poncu , Alexandru Tifrea , Elena Burceanu

A popular new method in mechanistic interpretability is to train high-dimensional sparse autoencoders (SAEs) on neuron activations and use SAE features as the atomic units of analysis. However, the body of evidence on whether SAE feature…

Machine Learning · Computer Science 2024-09-10 Maheep Chaudhary , Atticus Geiger

Sparse autoencoders (SAEs) are one of the main methods to interpret the inner workings of deep neural networks (DNNs), decomposing activations into higher-dimensional features. However, they exhibit critical shortcomings where a large…

Machine Learning · Computer Science 2026-05-19 Michał Brzozowski , Neo Christopher Chung

Sparse Autoencoders (SAEs) are an interpretability technique aimed at decomposing neural network activations into interpretable units. However, a major bottleneck for SAE development has been the lack of high-quality performance metrics,…

Machine Learning · Computer Science 2024-12-02 Adam Karvonen , Can Rager , Samuel Marks , Neel Nanda

Unsupervised approaches to large language model (LLM) interpretability, such as sparse autoencoders (SAEs), offer a way to decode LLM activations into interpretable and, ideally, controllable concepts. On the one hand, these approaches…

Machine Learning · Computer Science 2026-03-03 Shruti Joshi , Andrea Dittadi , Sébastien Lachapelle , Dhanya Sridhar

Sparse Autoencoders (SAEs) provide potentials for uncovering structured, human-interpretable representations in Large Language Models (LLMs), making them a crucial tool for transparent and controllable AI systems. We systematically analyze…

Machine Learning · Computer Science 2026-02-03 Jack Gallifant , Shan Chen , Kuleen Sasse , Hugo Aerts , Thomas Hartvigsen , Danielle S. Bitterman

Sparse autoencoders (SAEs) extract millions of interpretable features from a language model, but flat feature inventories aren't very useful on their own. Domain concepts get mixed with generic and weakly grounded features, while related…

Artificial Intelligence · Computer Science 2026-04-29 John Winnicki , Abeynaya Gnanasekaran , Eric Darve

Sparse Autoencoder (SAE) has emerged as a powerful tool for mechanistic interpretability of large language models. Recent works apply SAE to protein language models (PLMs), aiming to extract and analyze biologically meaningful features from…

Quantitative Methods · Quantitative Biology 2026-01-21 Xiangyu Liu , Haodi Lei , Yi Liu , Yang Liu , Wei Hu

Learning hierarchical features in Sparse Autoencoders (SAEs) is essential for capturing the structured nature of real-world data and mitigating issues like feature absorption or splitting. Existing works attempt to identify hierarchical…

Machine Learning · Computer Science 2026-05-12 Tue M. Cao , Hoang X. Nhat , Raed Alharbi , Phi Le Nguyen , My T. Thai

Sparse Autoencoders (SAEs) aim to decompose the activation space of large language models (LLMs) into human-interpretable latent directions or features. As we increase the number of features in the SAE, hierarchical features tend to split…

Computation and Language · Computer Science 2025-11-18 David Chanin , James Wilken-Smith , Tomáš Dulka , Hardik Bhatnagar , Satvik Golechha , Joseph Bloom

Sparse autoencoders (SAEs) are increasingly used for safety-relevant applications including alignment detection and model steering. These use cases require SAE latents to be as atomic as possible. Each latent should represent a single…

Machine Learning · Computer Science 2026-04-07 Matthew Levinson