mlsanityData to 14 Oct 2026

Sparse Autoencoder Features for Classifications and Transferability

arXiv 2502.11367 ↗First posted 17 Feb 2025cs.LG cs.AI cs.CLcited by 21

On the map: Mechanistic interpretability → SAE features and their applications

Abstract (arXiv)

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 SAE for interpretable feature extraction from LLMs in safety-critical classification tasks. Our framework evaluates (1) model-layer selection and scaling properties, (2) SAE architectural configurations, including width and pooling strategies, and (3) the effect of binarizing continuous SAE activations. SAE-derived features achieve macro F1 > 0.8, outperforming hidden-state and BoW baselines while demonstrating cross-model transfer from Gemma 2 2B to 9B-IT models. These features generalize in a zero-shot manner to cross-lingual toxicity detection and visual classification tasks. Our analysis highlights the significant impact of pooling strategies and binarization thresholds, showing that binarization offers an efficient alternative to traditional feature selection while maintaining or improving performance. These findings establish new best practices for SAE-based interpretability and enable scalable, transparent deployment of LLMs in real-world applications. Full repo: https://github.com/shan23chen/MOSAIC.

Key facts. Cited by 21 papers in the map (data through 14 Oct 2026). Placed in SAE features and their applications, one of 13 directions in Mechanistic interpretability. Builds on 25 of its 45 references that are in the map. First posted on arXiv 17 Feb 2025.

Builds on

25 of 45 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders28 Jan
  2. 2024Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering21 Oct
  3. 2024Improving Dictionary Learning with Gated Sparse Autoencoders24 Apr
  4. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Models15 Sep

Other references · 21

  1. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  2. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  3. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  4. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
  5. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,165
  6. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 1,032
  7. 2022Toy Models of Superpositioncited by 971
  8. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 435
  9. 2020Efficient Intent Detection with Dual Sentence Encoderscited by 433
  10. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 394
  11. 2021Entity-Based Knowledge Conflicts in Question Answeringcited by 356
  12. 2022The Alignment Problem from a Deep Learning Perspectivecited by 258
9 more
  1. 2023An Overview of Catastrophic AI Riskscited by 222
  2. 2023Attribution Patching Outperforms Automated Circuit Discoverycited by 209
  3. 2022MASSIVE: A 1M-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languagescited by 146
  4. 2024Looking Inward: Language Models Can Learn About Themselves by Introspectioncited by 113
  5. 2021An Interpretability Illusion for BERTcited by 97
  6. 2025Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languagescited by 59
  7. 2018Baseline Needs More Love: On Simple Word-Embedding-Based Models and Associated Pooling Mechanismscited by 53
  8. 2024An X-Ray Is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generationcited by 25
  9. 2024LLaVA-Gemma: Accelerating Multimodal Foundation Models with a Compact Language Modelcited by 14

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 10; 2026: 11.

21 papers.

Sort:

By area:

  1. 2025A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Modelscited by 76
  2. 2025SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Control of Language Modelscited by 29
  3. 2025SAE-V: Interpreting Multimodal Models for Enhanced Alignmentcited by 23
  4. 2025Investigating task-specific prompts and sparse autoencoders for activation monitoringcited by 15
  5. 2026When Benchmarks Lie: Evaluating Malicious Prompt Classifiers Under True Distribution Shiftcited by 11
  6. 2025Interpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM Safetycited by 10
  7. 2025A Comparative Analysis of Sparse Autoencoder and Activation Difference in Language Model Steeringcited by 4
  8. 2025Unveiling Decision-Making in LLMs for Text Classification : Extraction of influential and interpretable concepts with Sparse Autoencoderscited by 3
  9. 2025ProtSAE: Disentangling and Interpreting Protein Language Models via Semantically-Guided Sparse Autoencoderscited by 3
  10. 2025Binary Sparse Coding for Interpretabilitycited by 3
  11. 2026Towards Interpretable Hallucination Analysis and Mitigation in LVLMs via Contrastive Neuron Steeringcited by 3
  12. 2025Can SAEs reveal and mitigate racial biases of LLMs in healthcare?cited by 2
9 more
  1. 2026Beyond Activation Patterns: A Weight-Based Out-of-Context Explanation of Sparse Autoencoder Featurescited by 1
  2. 2026Measuring Monosemanticity in Sparse Autoencoders via Latent Activation Coherencecited by 1
  3. 2026MedConcept: Unsupervised Concept Discovery for Interpretability in Medical VLMscited by 0
  4. 2026Lost or Hidden? A Concept-Level Forgetting in Supervised Continual Learningcited by 0
  5. 2026Robust and Generalizable Safety Steering for Text-to-Image Diffusion Transformerscited by 0
  6. 2026Thinking vs. NoThinking: Towards Interpreting Reasoning Mechanisms of Large Language Models via Sparse Autoencoderscited by 0
  7. 2026Test-Time Unlearning via Sparse Autoencodercited by 0
  8. 2026Clinical Concept Centers in LLMscited by 0
  9. 2026Sparse Autoencoders for Mechanistic Interpretability in NLP: A Critical Survey of Recoverability, Interpretability, and Utilitycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability
  2. 2025Sparse Autoencoders Trained on the Same Data Learn Different Features
  3. 2025Measuring and Guiding Monosemanticity
  4. 2026Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models
  5. 2025AbsTopK: Rethinking Sparse Autoencoders For Bidirectional Features
  6. 2026Are Sparse Autoencoder Benchmarks Reliable?
  7. 2025Probing the Representational Power of Sparse Autoencoders in Vision Models
  8. 2025AdaptiveK: Complexity-Driven Sparse Autoencoders for Interpretable Language Model Representations