mlsanityData to 14 Oct 2026

Sparse Autoencoders for Interpretable Out-of-Distribution Detection

arXiv 2607.12094 ↗First posted 13 Jul 2026cs.LG cs.AIcited by 0

On the map: Mechanistic interpretability → Sparse autoencoder training and evaluation

Abstract (arXiv)

Reliable detection of out-of-distribution (OOD) samples is crucial for the safe deployment of machine learning models. Neural networks often produce overconfident predictions for inputs that deviate from their training data, leading to significant degradation in performance. While many OOD detection methods focus on the final output layer, they neglect the rich hierarchical information present in intermediate network layers. This paper introduces a novel approach that leverages sparse autoencoders (SAEs) to learn interpretable features from these intermediate activations. We find that in-distribution (ID) and OOD data activate distinct sets of these sparse features. We propose a new OOD score derived from the cosine similarity between the sparse feature activations of a test sample and the mean activations of ID classes. Our post-hoc detection method not only achieves state-of-the-art performance on standard OOD detection benchmarks, but yields interpretable insights into how distribution shift affects learned representations.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Sparse autoencoder training and evaluation, one of 13 directions in Mechanistic interpretability. Builds on 29 of its 39 references that are in the map. First posted on arXiv 13 Jul 2026.

Builds on

29 of 39 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 29 Aug

Other references · 28

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  4. 2017On Calibration of Modern Neural Networkscited by 4,960
  5. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,349
  6. 2018UMAP: Uniform Manifold Approximation and Projection for Dimension Reductioncited by 2,898
  7. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,877
  8. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,385
  9. 2013Describing Textures in the Wildcited by 2,125
  10. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,850
  11. 2018A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attackscited by 1,485
  12. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,420
16 more
  1. 2020Energy-based Out-of-distribution Detectioncited by 1,299
  2. 2021Generalized Out-of-Distribution Detection: A Surveycited by 802
  3. 2022Out-of-Distribution Detection with Deep Nearest Neighborscited by 600
  4. 2024Scaling and evaluating sparse autoencoderscited by 553
  5. 2019Scaling Out-of-Distribution Detection for Real-World Settingscited by 465
  6. 2022ViM: Out-Of-Distribution with Virtual-logit Matchingcited by 405
  7. 2021Open-Set Recognition: a Good Closed-Set Classifier is All You Need?cited by 353
  8. 2013k-Sparse Autoencoderscited by 215
  9. 2022Extremely Simple Activation Shaping for Out-of-Distribution Detectioncited by 212
  10. 2021A Simple Fix to Mahalanobis Distance for Improving Near-OOD Detectioncited by 207
  11. 2025Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignmentcited by 62
  12. 2020Towards Understanding Hierarchical Learning: Benefits of Neural Representationscited by 46
  13. 2024Unexplored Faces of Robustness and Out-of-Distribution: Covariate Shifts in Environment and Sensor Domainscited by 22
  14. 2022Concept-based Explanations for Out-Of-Distribution Detectorscited by 16
  15. 2019Unsupervised Boosting-based Autoencoder Ensembles for Outlier Detectioncited by 8
  16. 2025Leveraging Intermediate Representations for Better Out-of-Distribution Detectioncited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Sparsity as a Key: Unlocking New Insights from Latent Structures for Out-of-Distribution Detection
  2. 2026When Confidence Lacks Concepts: Interpretable OOD Detection via Representation Perturbations
  3. 2026Improving Robustness In Sparse Autoencoders via Masked Regularization
  4. 2024NODI: Out-Of-Distribution Detection with Noise from Diffusion
  5. 2025Beyond Binary Out-of-Distribution Detection: Characterizing Distributional Shifts with Multi-Statistic Diffusion Trajectories
  6. 2024Enhancing OOD Detection Using Latent Diffusion
  7. 2023Class Relevance Learning For Out-of-distribution Detection
  8. 2024Non-Linear Outlier Synthesis for Out-of-Distribution Detection