mlsanityData to 7 Oct 2026

Why Normalizing Flows Fail to Detect Out-of-Distribution Data

arXiv 2006.08545 ↗First posted 15 Jun 2020stat.ML cs.LGcited by 220

Not on the map. Why?

Abstract (arXiv)

Detecting out-of-distribution (OOD) data is crucial for robust machine learning systems. Normalizing flows are flexible deep generative models that often surprisingly fail to distinguish between in- and out-of-distribution data: a flow trained on pictures of clothing assigns higher likelihood to handwritten digits. We investigate why normalizing flows perform poorly for OOD detection. We demonstrate that flows learn local pixel correlations and generic image-to-latent-space transformations which are not specific to the target image dataset. We show that by modifying the architecture of flow coupling layers we can bias the flow towards learning the semantic structure of the target data, improving OOD detection. Our investigation reveals that properties that enable flows to generate high-fidelity images can have a detrimental effect on OOD detection.

Key facts. Cited by 220 papers in the map (data through 7 Oct 2026). First posted on arXiv 15 Jun 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 36; 2022: 38; 2023: 41; 2024: 43; 2025: 29; 2026: 33.

220 papers.

Sort:

By area:

  1. 2021Generalized Out-of-Distribution Detection: A Surveycited by 796
  2. 2021CutPaste: Self-Supervised Learning for Anomaly Detection and Localizationcited by 529
  3. 2021ReAct: Out-of-distribution Detection With Rectified Activationscited by 486
  4. 2021On the Importance of Gradients for Detecting Distributional Shifts in the Wildcited by 338
  5. 2021CFLOW-AD: Real-Time Unsupervised Anomaly Detection with Localization via Conditional Normalizing Flowscited by 320
  6. 2021Offline Reinforcement Learning with Fisher Divergence Critic Regularizationcited by 237
  7. 2022Delving into Out-of-Distribution Detection with Vision-Language Representationscited by 215
  8. 2021Low-Light Image Enhancement with Normalizing Flowcited by 184
  9. 2023OpenOOD v1.5: Enhanced Benchmark for Out-of-Distribution Detectioncited by 178
  10. 2022How to Exploit Hyperspherical Embeddings for Out-of-Distribution Detection?cited by 128
  11. 2023Score-Based Diffusion Models as Principled Priors for Inverse Imagingcited by 126
  12. 2021Fully Convolutional Cross-Scale-Flows for Image-based Defect Detectioncited by 117
38 more
  1. 2023LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learningcited by 107
  2. 2021DEUP: Direct Epistemic Uncertainty Predictioncited by 102
  3. 2021Understanding Failures in Out-of-Distribution Detection with Deep Generative Modelscited by 89
  4. 2021Types of Out-of-Distribution Texts and How to Detect Themcited by 80
  5. 2021Graph Posterior Network: Bayesian Predictive Uncertainty for Node Classificationcited by 79
  6. 2023PyramidFlow: High-Resolution Defect Contrastive Localization using Pyramid Normalizing Flowcited by 74
  7. 2022Unknown-Aware Object Detection: Learning What You Don't Know from Videos in the Wildcited by 70
  8. 2021On the Impact of Spurious Correlation for Out-of-distribution Detectioncited by 68
  9. 2022Normalizing Flows for Human Pose Anomaly Detectioncited by 59
  10. 2023Unsupervised Out-of-Distribution Detection with Diffusion Inpaintingcited by 57
  11. 2022RankFeat: Rank-1 Feature Removal for Out-of-distribution Detectioncited by 52
  12. 2022Boosting Out-of-distribution Detection with Typical Featurescited by 48
  13. 2024A Closer Look at the Robustness of Contrastive Language-Image Pre-Training (CLIP)cited by 44
  14. 2023On Advantages of Mask-level Recognition for Outlier-aware Segmentationcited by 42
  15. 2024A Geometric View of Data Complexity: Efficient Local Intrinsic Dimension Estimation with Diffusion Modelscited by 41
  16. 2023GL-MCM: Global and Local Maximum Concept Matching for Zero-Shot Out-of-Distribution Detectioncited by 36
  17. 2023LINe: Out-of-Distribution Detection by Leveraging Important Neuronscited by 34
  18. 2024Outlier detection by ensembling uncertainty with negative objectnesscited by 34
  19. 2021Autoencoders for unsupervised anomaly detection in high energy physicscited by 32
  20. 2021On the Out-of-distribution Generalization of Probabilistic Image Modellingcited by 32
  21. 2022Diagnosing and Fixing Manifold Overfitting in Deep Generative Modelscited by 28
  22. 2022Watermarking for Out-of-distribution Detectioncited by 28
  23. 2023Neuron Activation Coverage: Rethinking Out-of-distribution Detection and Generalizationcited by 28
  24. 2024Hierarchical Gaussian Mixture Normalizing Flow Modeling for Unified Anomaly Detectioncited by 28
  25. 2023DiffGuard: Semantic Mismatch-Guided Out-of-Distribution Detection using Pre-trained Diffusion Modelscited by 25
  26. 2023MSFlow: Multi-Scale Flow-based Framework for Unsupervised Anomaly Detectioncited by 25
  27. 2022Out-of-distribution Detection via Frequency-regularized Generative Modelscited by 23
  28. 2021On the Properties of Kullback-Leibler Divergence Between Multivariate Gaussian Distributionscited by 22
  29. 2022PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularizationcited by 22
  30. 2022Out-of-Distribution Detection with Semantic Mismatch under Maskingcited by 22
  31. 2024A Geometric Explanation of the Likelihood OOD Detection Paradoxcited by 22
  32. 2022Rainproof: An Umbrella To Shield Text Generators From Out-Of-Distribution Datacited by 21
  33. 2024MULDE: Multiscale Log-Density Estimation via Denoising Score Matching for Video Anomaly Detectioncited by 21
  34. 2024Your Data Is Not Perfect: Towards Cross-Domain Out-of-Distribution Detection in Class-Imbalanced Datacited by 21
  35. 2022Igeood: An Information Geometry Approach to Out-of-Distribution Detectioncited by 20
  36. 2022Improving the Intra-class Long-tail in 3D Detection via Rare Example Miningcited by 19
  37. 2024Diffusion Models as Cartoonists: The Curious Case of High Density Regionscited by 19
  38. 2022Data Redaction from Pre-trained GANscited by 18

50 most cited of 220 All 220 (JSON)