mlsanityData to 14 Oct 2026

Is Out-of-Distribution Detection Learnable?

arXiv 2210.14707 ↗First posted 26 Oct 2022cs.LG stat.MLcited by 109

Not on the map. Why?

Abstract (arXiv)

Supervised learning aims to train a classifier under the assumption that training and test data are from the same distribution. To ease the above assumption, researchers have studied a more realistic setting: out-of-distribution (OOD) detection, where test data may come from classes that are unknown during training (i.e., OOD data). Due to the unavailability and diversity of OOD data, good generalization ability is crucial for effective OOD detection algorithms. To study the generalization of OOD detection, in this paper, we investigate the probably approximately correct (PAC) learning theory of OOD detection, which is proposed by researchers as an open problem. First, we find a necessary condition for the learnability of OOD detection. Then, using this condition, we prove several impossibility theorems for the learnability of OOD detection under some scenarios. Although the impossibility theorems are frustrating, we find that some conditions of these impossibility theorems may not hold in some practical scenarios. Based on this observation, we next give several necessary and sufficient conditions to characterize the learnability of OOD detection in some practical scenarios. Lastly, we also offer theoretical supports for several representative OOD detection works based on our OOD theory.

Key facts. Cited by 109 papers in the map (data through 14 Oct 2026). Builds on 56 of its 99 references that are in the map. First posted on arXiv 26 Oct 2022.

Builds on

56 of 99 references.

References · 56

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2016Densely Connected Convolutional Networkscited by 6,720
  4. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,385
  5. 2018Glow: Generative Flow with Invertible 1x1 Convolutionscited by 1,545
  6. 2018A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attackscited by 1,485
  7. 2017Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networkscited by 1,328
  8. 2020Energy-based Out-of-distribution Detectioncited by 1,299
  9. 2018Deep Anomaly Detection with Outlier Exposurecited by 942
  10. 2021Generalized Out-of-Distribution Detection: A Surveycited by 802
  11. 2022Out-of-Distribution Detection with Deep Nearest Neighborscited by 600
  12. 2015Towards Open Set Deep Networkscited by 538
38 more
  1. 2021ReAct: Out-of-distribution Detection With Rectified Activationscited by 486
  2. 2019Likelihood Ratios for Out-of-Distribution Detectioncited by 453
  3. 2018Do Deep Generative Models Know What They Don't Know?cited by 443
  4. 2020Generalized ODIN: Detecting Out-of-distribution Image without Learning from Out-of-distribution Datacited by 408
  5. 2021On the Importance of Gradients for Detecting Distributional Shifts in the Wildcited by 339
  6. 2017Nearly-tight VC-dimension and pseudodimension bounds for piecewise linear neural networkscited by 311
  7. 2021Exploring the Limits of Out-of-Distribution Detectioncited by 283
  8. 2017Positive-Unlabeled Learning with Non-Negative Risk Estimatorcited by 219
  9. 2021A Simple Fix to Mahalanobis Distance for Improving Near-OOD Detectioncited by 207
  10. 2018Reducing Network Agnostophobiacited by 165
  11. 2021A Unified Survey on Anomaly, Novelty, Open-Set, and Out-of-Distribution Detection: Solutions and Future Challengescited by 154
  12. 2021Adversarial Reciprocal Points Learning for Open Set Recognitioncited by 153
  13. 2020Likelihood Regret: An Out-of-Distribution Detection Score For Variational Auto-encodercited by 145
  14. 2020DROCC: Deep Robust One-Class Classificationcited by 121
  15. 2020ATOM: Robustifying Out-of-distribution Detection Using Outlier Miningcited by 116
  16. 2020Learning Open Set Network with Discriminative Reciprocal Pointscited by 115
  17. 2019Recent Advances in Algorithmic High-Dimensional Robust Statisticscited by 101
  18. 2021Can multi-label classification networks know what they don't know?cited by 100
  19. 2018Generative Probabilistic Novelty Detection with Adversarial Autoencoderscited by 99
  20. 2021Evidential Deep Learning for Open Set Action Recognitioncited by 96
  21. 2021MOOD: Multi-level Out-of-distribution Detectioncited by 94
  22. 2021Understanding Failures in Out-of-Distribution Detection with Deep Generative Modelscited by 89
  23. 2021Provable Guarantees for Understanding Out-of-distribution Detectioncited by 78
  24. 2021On the Impact of Spurious Correlation for Out-of-distribution Detectioncited by 68
  25. 2022Full-Spectrum Out-of-Distribution Detectioncited by 63
  26. 2021Optimal strategies for reject option classifierscited by 61
  27. 2016Depth-Width Tradeoffs in Approximating Natural Functions with Neural Networkscited by 59
  28. 2020Robust Out-of-distribution Detection for Neural Networkscited by 57
  29. 2020Progressive Graph Learning for Open-Set Domain Adaptationcited by 56
  30. 2020Classification with Rejection Based on Cost-sensitive Classificationcited by 55
  31. 2020What Can Be Transferred: Unsupervised Domain Adaptation for Endoscopic Lesions Segmentationcited by 50
  32. 2020Outlier Robust Mean Estimation with Subgaussian Rates via Stabilitycited by 50
  33. 2019Out-of-distribution Detection in Classifiers via Generationcited by 49
  34. 2019Open Set Domain Adaptation: Theoretical Bound and Algorithmcited by 48
  35. 2020Beyond Perturbations: Learning Guarantees with Arbitrary Adversarial Test Examplescited by 44
  36. 2019On the Calibration of Multiclass Classification with Rejectioncited by 41
  37. 2018Open Category Detection with PAC Guaranteescited by 35
  38. 2021Learning Bounds for Open-Set Learningcited by 27

50 most cited of 56 All 56 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 3; 2023: 38; 2024: 31; 2025: 28; 2026: 9.

109 papers.

Sort:

By area:

  1. 2022Delving into Out-of-Distribution Detection with Vision-Language Representationscited by 216
  2. 2024ARGS: Alignment as Reward-Guided Searchcited by 103
  3. 2023Provable Dynamic Fusion for Low-Quality Multimodal Datacited by 82
  4. 2024Process Reward Model with Q-Value Rankingscited by 80
  5. 2023Dream the Impossible: Outlier Imagination with Diffusion Modelscited by 77
  6. 2023Nearest Neighbor Guidance for Out-of-Distribution Detectioncited by 64
  7. 2023FedFed: Feature Distillation against Data Heterogeneity in Federated Learningcited by 54
  8. 2023Out-of-distribution Detection with Implicit Outlier Transformationcited by 43
  9. 2023Diversified Outlier Exposure for Out-of-Distribution Detection via Informative Extrapolationcited by 38
  10. 2023Learning to Augment Distributions for Out-of-Distribution Detectioncited by 38
  11. 2023Federated Incremental Semantic Segmentationcited by 36
  12. 2024How Does Unlabeled Data Provably Help Out-of-Distribution Detection?cited by 36
38 more
  1. 2023Maskomaly:Zero-Shot Mask Anomaly Segmentationcited by 34
  2. 2024Out-of-Domain Generalization in Dynamical Systems Reconstructioncited by 33
  3. 2023A Survey on Out-of-Distribution Detection in NLPcited by 31
  4. 2023Out-of-distribution Detection Learning with Unreliable Out-of-distribution Sourcescited by 31
  5. 2023Learnability and Algorithm for Continual Learningcited by 28
  6. 2023Understanding the Feature Norm for Out-of-Distribution Detectioncited by 27
  7. 2024Out-of-Distribution Detection: A Task-Oriented Survey of Recent Advancescited by 26
  8. 2023DIVERSIFY: A General Framework for Time Series Out-of-distribution Detection and Generalizationcited by 24
  9. 2024Kernel PCA for Out-of-Distribution Detectioncited by 24
  10. 2024ConjNorm: Tractable Density Estimation for Out-of-Distribution Detectioncited by 22
  11. 2023Unleashing Mask: Explore the Intrinsic Out-of-Distribution Detection Capabilitycited by 16
  12. 2023Unsupervised Layer-wise Score Aggregation for Textual OOD Detectioncited by 15
  13. 2023Hard Sample Matters a Lot in Zero-Shot Quantizationcited by 15
  14. 2023Heterogeneous Forgetting Compensation for Class-Incremental Learningcited by 15
  15. 2023Average of Pruning: Improving Performance and Stability of Out-of-Distribution Detectioncited by 14
  16. 2025OpenMIBOOD: Open Medical Imaging Benchmarks for Out-Of-Distribution Detectioncited by 14
  17. 2025Handling Out-of-Distribution Data: A Surveycited by 14
  18. 2024NoiseDiffusion: Correcting Noise for Image Interpolation with Diffusion Models beyond Spherical Linear Interpolationcited by 13
  19. 2024Conjugated Semantic Pool Improves OOD Detection with Pre-trained Vision-Language Modelscited by 13
  20. 2023Domain Generalization via Nuclear Norm Regularizationcited by 12
  21. 2023On the Powerfulness of Textual Outlier Exposure for Visual OoD Detectioncited by 12
  22. 2024Hard to Explain: On the Computational Hardness of In-Distribution Model Interpretationcited by 11
  23. 2025Do we really have to filter out random noise in pre-training data for language models?cited by 11
  24. 2023Universal Domain Adaptation from Foundation Models: A Baseline Studycited by 10
  25. 2023Meta OOD Learning for Continuously Adaptive OOD Detectioncited by 10
  26. 2023Revisiting Deep Ensemble for Out-of-Distribution Detection: A Loss Landscape Perspectivecited by 10
  27. 2024Revisiting Confidence Estimation: Towards Reliable Failure Predictioncited by 10
  28. 2024LEGO-Learn: Label-Efficient Graph Open-Set Learningcited by 10
  29. 2025Compositional Generalization from Learned Skills via CoT Training: A Theoretical and Structural Analysis for Reasoningcited by 10
  30. 2025Noisy Test-Time Adaptation in Vision-Language Modelscited by 10
  31. 2025On Continuous Monitoring of Risk Violations under Unknown Shiftcited by 10
  32. 2023Model-free Test Time Adaptation for Out-Of-Distribution Detectioncited by 9
  33. 2023Moderately Distributional Exploration for Domain Generalizationcited by 8
  34. 2023Unlocking the Power of Open Set : A New Perspective for Open-Set Noisy Label Learningcited by 8
  35. 2023Novelty Detection in Reinforcement Learning with World Modelscited by 8
  36. 2023Graph Neural Networks for Pressure Estimation in Water Distribution Systemscited by 8
  37. 2024On the Learnability of Out-of-distribution Detectioncited by 8
  38. 2024Adaptive Hierarchical Graph Cut for Multi-granularity Out-of-distribution Detectioncited by 8

50 most cited of 109 All 109 (JSON)