mlsanityData to 7 Oct 2026

Improving Dictionary Learning with Gated Sparse Autoencoders

arXiv 2404.16014 ↗First posted 24 Apr 2024cs.LG cs.AIcited by 171

Not on the map. Why?

Abstract (arXiv)

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 Gated SAEs is to separate the functionality of (a) determining which directions to use and (b) estimating the magnitudes of those directions: this enables us to apply the L1 penalty only to the former, limiting the scope of undesirable side effects. Through training SAEs on LMs of up to 7B parameters we find that, in typical hyper-parameter ranges, Gated SAEs solve shrinkage, are similarly interpretable, and require half as many firing features to achieve comparable reconstruction fidelity.

Key facts. Cited by 171 papers in the map (data through 7 Oct 2026). Builds on 19 of its 51 references that are in the map. First posted on arXiv 24 Apr 2024.

Builds on

19 of 51 references.

References · 19

  1. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  2. 2023Pythia: A Suite for Analyzing Large Language Models Across Training and Scalingcited by 1,734
  3. 2020GLU Variants Improve Transformercited by 1,724
  4. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  5. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,130
  6. 2022Toy Models of Superpositioncited by 928
  7. 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 858
  8. 2023Progress measures for grokking via mechanistic interpretabilitycited by 822
  9. 2016Language Modeling with Gated Convolutional Networkscited by 688
  10. 2023The Linear Representation Hypothesis and the Geometry of Large Language Modelscited by 679
  11. 2018Disentangling by Factorisingcited by 614
  12. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
7 more
  1. 2023Finding Neurons in a Haystack: Case Studies with Sparse Probingcited by 328
  2. 2023Linear Representations of Sentiment in Large Language Modelscited by 157
  3. 2021An Interpretability Illusion for BERTcited by 94
  4. 2013Deep Learning of Representations: Looking Forwardcited by 84
  5. 2023Copy Suppression: Comprehensively Understanding an Attention Headcited by 73
  6. 2023Codebook Features: Sparse and Discrete Interpretability for Neural Networkscited by 45
  7. 2019JumpReLU: A Retrofit Defense Strategy for Adversarial Attackscited by 14

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 39; 2025: 82; 2026: 50.

171 papers.

Sort:

By area:

  1. 2024Scaling and evaluating sparse autoencoderscited by 530
  2. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 381
  3. 2024Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoderscited by 283
  4. 2024Transcoders Find Interpretable LLM Feature Circuitscited by 186
  5. 2025Learning Multi-Level Features with Matryoshka Sparse Autoencoderscited by 136
  6. 2024BatchTopK Sparse Autoencoderscited by 126
  7. 2025SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretabilitycited by 118
  8. 2024A Practical Review of Mechanistic Interpretability for Transformer-Based Language Modelscited by 113
  9. 2025Are Sparse Autoencoders Useful? A Case Study in Sparse Probingcited by 109
  10. 2024Steering Language Model Refusal with Sparse Autoencoderscited by 89
  11. 2024A Primer on the Inner Workings of Transformer-based Language Modelscited by 83
  12. 2025Persona Features Control Emergent Misalignmentcited by 76
38 more
  1. 2025A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Modelscited by 73
  2. 2024Sparse autoencoders reveal selective remapping of visual concepts during adaptationcited by 66
  3. 2024Identifying Functionally Important Features with End-to-End Sparse Dictionary Learningcited by 64
  4. 2025Projecting Assumptions: The Duality Between Sparse Autoencoders and Concept Geometrycited by 58
  5. 2025Interpreting CLIP with Hierarchical Sparse Autoencoderscited by 54
  6. 2024Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Modelscited by 51
  7. 2025From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuitcited by 45
  8. 2024Interpreting Attention Layer Outputs with Sparse Autoencoderscited by 43
  9. 2024Talking Heads: Understanding Inter-layer Communication in Transformer Language Modelscited by 39
  10. 2024Efficient Dictionary Learning with Switch Sparse Autoencoderscited by 38
  11. 2024Interpreting the Second-Order Effects of Neurons in CLIPcited by 36
  12. 2025Understanding the Repeat Curse in Large Language Models from a Feature Perspectivecited by 36
  13. 2024Mathematical Models of Computation in Superpositioncited by 34
  14. 2024Towards Universality: Studying Mechanistic Similarity Across Language Model Architecturescited by 31
  15. 2024Towards Unifying Interpretability and Control: Evaluation via Interventioncited by 28
  16. 2024Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoderscited by 25
  17. 2024Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoderscited by 25
  18. 2025Route Sparse Autoencoder to Interpret Large Language Modelscited by 25
  19. 2024InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniquescited by 24
  20. 2024The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysiscited by 24
  21. 2024Relational Composition in Neural Networks: A Survey and Call to Actioncited by 23
  22. 2024An X-Ray Is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generationcited by 23
  23. 2024Automatically Identifying Local and Global Circuits with Linear Computation Graphscited by 22
  24. 2026Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Modelscited by 22
  25. 2025Prisma: An Open Source Toolkit for Mechanistic Interpretability in Vision and Videocited by 20
  26. 2024Disentangling Dense Embeddings with Sparse Autoencoderscited by 19
  27. 2025Sparse Autoencoder Features for Classifications and Transferabilitycited by 19
  28. 2024Towards Inference-time Category-wise Safety Steering for Large Language Modelscited by 17
  29. 2024Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoderscited by 17
  30. 2025Rethinking Evaluation of Sparse Autoencoders through the Representation of Polysemous Wordscited by 17
  31. 2024Analyzing (In)Abilities of SAEs via Formal Languagescited by 16
  32. 2025From superposition to sparse codes: interpretable representations in neural networkscited by 16
  33. 2024Decoding Dark Matter: Specialized Sparse Autoencoders for Interpreting Rare Concepts in Foundation Modelscited by 15
  34. 2025Analyzing Finetuning Representation Shift for Multimodal LLMs Steeringcited by 15
  35. 2024Learning biologically relevant features in a pathology foundation model using sparse autoencoderscited by 14
  36. 2025Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEscited by 14
  37. 2025LinguaLens: Towards Interpreting Linguistic Mechanisms of Large Language Models via Sparse Auto-Encodercited by 13
  38. 2025I Predict Therefore I Am: Is Next Token Prediction Enough to Learn Human-Interpretable Concepts from Data?cited by 13

50 most cited of 171 All 171 (JSON)