Sparse Autoencoders

Also known as SAE

Sparse autoencoders learn a dictionary of features that can reconstruct a model’s dense activations using only a few active features at a time, helping researchers investigate superposition in language, vision, and multimodal models. Common approaches include L1- and Top-k-based training, BatchTopK, and structured variants that model feature groups, hierarchies, or relationships across layers and modalities. Researchers use the resulting features to explain model behavior and to steer, audit, or remove targeted behaviors, while testing them with reconstruction and interpretability benchmarks as well as causal interventions. A central open tension is that sparse, human-readable features are not necessarily stable across training runs or faithful causes of behavior, and better reconstruction does not by itself guarantee reliable explanations or control.

Latest papers

220 papers42 papers in the last four weeks, up 133% on the four weeks before. 0.2% of all new papers.

Beyond Dense States: Sparse Transcoders as Causally Testable Operators for LLM Latent Reasoning
Feb 2, 2026Yadong Wang, Haodong Chen, Yu Tian +3Sparse Autoencoders Continuous Latent Reasoning
Nanjing University of Aeronautics and Astronautics · State Key Laboratory of Ocean Sensing, ZJU-Hangzhou Global Scientific and Technological Innovation Center, Zhejiang University, Hangzhou, 311215, China · Security Capability Center
Mechanistic Knobs in LLMs: Retrieving and Steering High-Order Semantic Features via Sparse Autoencoders
Jan 6, 2026Ruikang Zhang, Shuo Wang, Qi SuLLM Interpretability Activation Steering
Peking University, Beijing, China
AlignSAE: Concept-Aligned Sparse Autoencoders
Dec 1, 2025Minglai Yang, Xinyu Guo, Zhengliang Shi +4Sparse Autoencoders LLM Interpretability
University of Arizona · Shandong University · LMU Munich +1
Towards Open-Ended Visual Scientific Discovery with Sparse Autoencoders
Nov 21, 2025Jacob Beattie, Samuel Stevens, Neil Rosser +2Vision Foundation Models Sparse Autoencoders
The Ohio State University · The University of Miami
Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders
Aug 22, 2025David Chanin, Adrià Garriga-AlonsoActivation Sparsity LLM Interpretability
University College London · MATS Research
Sparse Autoencoders Can Capture Language-Specific Concepts Across Diverse Languages
Jul 15, 2025Lyzander Marciano Andrylie, Inaya Rahmanisa, Mahardika Krisna Ihsani +3Multilingual Language Models Representation Learning
Faculty of Computer Science, Universitas Indonesia · Department of Natural Language Processing, MBZUAI
Position: Use Sparse Autoencoders to Discover Unknowns
Jun 30, 2025Kenny Peng, Rajiv Movva, Jon Kleinberg +2Explainable Artificial Intelligence Sparse Autoencoders
1Cornell University · 2UC Berkeley
Cross-Layer Discrete Concept Discovery for Interpreting Language Models
Jun 24, 2025Ankur Garg, Xuemin Yu, Hassan Sajjad +1Variational Autoencoders LLM Interpretability
Kronecker Factorization Improves Efficiency and Interpretability of Sparse Autoencoders
May 28, 2025Vadim Kurochkin, Yaroslav Aksenov, Daniil Laptev +2Representation Learning Activation Sparsity
T-Tech
SAEs Can Improve Unlearning: Dynamic Sparse Autoencoder Guardrails for Precision Unlearning in LLMs
Apr 11, 2025Aashiq Muhamed, Jacopo Bonato, Mona Diab +1Machine Unlearning LLM Unlearning
Carnegie Mellon University · Prometeia S.p.A.

Feb 2, 2026J/K move · Enter open · S save

Beyond Dense States: Sparse Transcoders as Causally Testable Operators for LLM Latent Reasoning

Yadong Wang, Haodong Chen, Yu Tian, Chuanxing Geng+2

Nanjing University of Aeronautics and Astronautics · State Key Laboratory of Ocean Sensing, ZJU-Hangzhou Global Scientific and Technological Innovation Center, Zhejiang University, Hangzhou, 311215, China · Security Capability Center

Sparse Autoencoders Continuous Latent Reasoning Causal Interventions in Language Models LLM Reasoning Latent Reasoning Latent CoT Reasoning

Open paper AskPDF arXiv

Latent reasoning reduces the token-generation cost of chain-of-thought reasoning by replacing explicit intermediate tokens with continuous latent transitions. However, existing latent reasoning methods usually rely on dense and entangled transitions, making their reasoning trajectories difficult to inspect or intervene on. We introduce LSTR (Latent Sparse Transcoder Reasoning), a framework that turns sparse transcoders from post-hoc diagnostic tools into in-loop, intervenable transition components for latent reasoning. At each latent step, a Latent Transition Transcoder (LTT) combines a linear skip path with a Top-k sparse innovation path, exposing a small set of active sparse features. Under matched compression settings, LSTR offers a mechanistically inspectable alternative to dense latent reasoning. On GSM8K-Aug, ablating only a few top-active sparse features reduces accuracy by up to 16.5%, whereas analogous interventions have much smaller effects in dense latent baselines. These results indicate that the active sparse features are causally involved in the latent transition process, rather than merely post-hoc descriptors. Additional experiments on mathematical benchmarks and StrategyQA suggest that sparse latent transitions can preserve the compression benefits of latent reasoning while making the resulting trajectories more inspectable and intervenable.