mlsanityData to 7 Oct 2026

Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift

arXiv 1810.11953 ↗First posted 29 Oct 2018stat.ML cs.LGcited by 240

Not on the map. Why?

Abstract (arXiv)

We might hope that when faced with unexpected inputs, well-designed software systems would fire off warnings. Machine learning (ML) systems, however, which depend strongly on properties of their inputs (e.g. the i.i.d. assumption), tend to fail silently. This paper explores the problem of building ML systems that fail loudly, investigating methods for detecting dataset shift, identifying exemplars that most typify the shift, and quantifying shift malignancy. We focus on several datasets and various perturbations to both covariates and label distributions with varying magnitudes and fractions of data affected. Interestingly, we show that across the dataset shifts that we explore, a two-sample-testing-based approach, using pre-trained classifiers for dimensionality reduction, performs best. Moreover, we demonstrate that domain-discriminating approaches tend to be helpful for characterizing shifts qualitatively and determining if they are harmful.

Key facts. Cited by 240 papers in the map (data through 7 Oct 2026). First posted on arXiv 29 Oct 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 30; 2022: 50; 2023: 33; 2024: 37; 2025: 47; 2026: 43.

240 papers.

Sort:

By area:

  1. 2021Towards Causal Representation Learningcited by 207
  2. 2021Trustworthy AI: From Principles to Practicescited by 206
  3. 2021MultiBench: Multiscale Benchmarks for Multimodal Representation Learningcited by 143
  4. 2022Leveraging Unlabeled Data to Predict Out-of-Distribution Performancecited by 141
  5. 2021Towards Robust and Reliable Algorithmic Recoursecited by 101
  6. 2022Distilling Model Failures as Directions in Latent Spacecited by 82
  7. 2021MMD Aggregated Two-Sample Testcited by 74
  8. 2021Mixture Proportion Estimation and PU Learning: A Modern Approachcited by 52
  9. 2022Diagnosing failures of fairness transfer across distribution shift in real-world medical settingscited by 45
  10. 2021GitTables: A Large-Scale Corpus of Relational Tablescited by 44
  11. 2021Neural Mean Discrepancy for Efficient Out-of-Distribution Detectioncited by 42
  12. 2023Look Before You Leap: An Exploratory Study of Uncertainty Measurement for Large Language Modelscited by 40
38 more
  1. 2022A survey on learning from imbalanced data streams: taxonomy, challenges, empirical study, and reproducible experimental frameworkcited by 39
  2. 2021Tracking the risk of a deployed model and detecting harmful distribution shiftscited by 38
  3. 2021How I failed machine learning in medical imaging -- shortcomings and recommendationscited by 37
  4. 2022Dataset Inference for Self-Supervised Modelscited by 36
  5. 2022A Validity Perspective on Evaluating the Justified Use of Data-driven Decision-making Algorithmscited by 30
  6. 2022Calibrated Selective Classificationcited by 30
  7. 2021Conditional Generative Models for Counterfactual Explanationscited by 28
  8. 2022Towards Explaining Distribution Shiftscited by 26
  9. 2022Operationalizing Machine Learning: An Interview Studycited by 25
  10. 2021Feature Shift Detection: Localizing Which Features Have Shifted via Conditional Distribution Testscited by 24
  11. 2022Linking Neural Collapse and L2 Normalization with Improved Out-of-Distribution Detection in Deep Neural Networkscited by 23
  12. 2023Calibration of Time-Series Forecasting: Detecting and Adapting Context-Driven Distribution Shiftcited by 23
  13. 2024A Geometric Explanation of the Likelihood OOD Detection Paradoxcited by 22
  14. 2022AutoML Two-Sample Testcited by 19
  15. 2022Estimating and Explaining Model Performance When Both Covariates and Labels Shiftcited by 19
  16. 2022Deconfounded Representation Similarity for Comparison of Neural Networkscited by 18
  17. 2022"Why did the Model Fail?": Attributing Model Performance Changes to Distribution Shiftscited by 18
  18. 2022A Learning Based Hypothesis Test for Harmful Covariate Shiftcited by 18
  19. 2021Uncertainty Estimation and Out-of-Distribution Detection for Counterfactual Explanations: Pitfalls and Solutionscited by 16
  20. 2024Unsupervised Concept Drift Detection from Deep Learning Representations in Real-timecited by 16
  21. 2021Model-agnostic and Scalable Counterfactual Explanations via Reinforcement Learningcited by 15
  22. 2021Newer is not always better: Rethinking transferability metrics, their peculiarities, stability and performancecited by 15
  23. 2024Public Domain 12M: A Highly Aesthetic Image-Text Dataset with Novel Governance Mechanismscited by 15
  24. 2021MLDemon: Deployment Monitoring for Machine Learning Systemscited by 14
  25. 2022GSCLIP : A Framework for Explaining Distribution Shifts in Natural Languagecited by 14
  26. 2022DC-Check: A Data-Centric AI checklist to guide the development of reliable machine learning systemscited by 14
  27. 2022PAC-Bayesian Domain Adaptation Bounds for Multiclass Learnerscited by 13
  28. 2022The Change that Matters in Discourse Parsing: Estimating the Impact of Domain Shift on Parser Errorcited by 12
  29. 2024Selective Classification Under Distribution Shiftscited by 12
  30. 2024Online Drift Detection with Maximum Concept Discrepancycited by 12
  31. 2022Active Learning Over Multiple Domains in Natural Language Taskscited by 11
  32. 2023OPORP: One Permutation + One Random Projectioncited by 11
  33. 2022Monitoring Model Deterioration with Explainable Uncertainty Estimation via Non-parametric Bootstrapcited by 9
  34. 2022Context-Aware Drift Detectioncited by 9
  35. 2023Endogenous Macrodynamics in Algorithmic Recoursecited by 9
  36. 2024Estimating Model Performance Under Covariate Shift Without Labelscited by 9
  37. 2025Avoiding Leakage Poisoning: Concept Interventions Under Distribution Shiftscited by 9
  38. 2022Suitability of Different Metric Choices for Concept Drift Detectioncited by 8

50 most cited of 240 All 240 (JSON)