mlsanityData to 14 Oct 2026

On Continuous Monitoring of Risk Violations under Unknown Shift

arXiv 2506.16416 ↗First posted 19 Jun 2025stat.ML cs.LGcited by 10

Not on the map. Why?

Abstract (arXiv)

Machine learning systems deployed in the real world must operate under dynamic and often unpredictable distribution shifts. This challenges the validity of statistical safety assurances on the system's risk established beforehand. Common risk control frameworks rely on fixed assumptions and lack mechanisms to continuously monitor deployment reliability. In this work, we propose a general framework for the real-time monitoring of risk violations in evolving data streams. Leveraging the 'testing by betting' paradigm, we propose a sequential hypothesis testing procedure to detect violations of bounded risks associated with the model's decision-making mechanism, while ensuring control on the false alarm rate. Our method operates under minimal assumptions on the nature of encountered shifts, rendering it broadly applicable. We illustrate the effectiveness of our approach by monitoring risks in outlier detection and set prediction under a variety of shifts.

Key facts. Cited by 10 papers in the map (data through 14 Oct 2026). Builds on 48 of its 60 references that are in the map. First posted on arXiv 19 Jun 2025.

Builds on

48 of 60 references.

References · 48

  1. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,375
  2. 2023A Comprehensive Survey of Continual Learning: Theory, Method and Applicationcited by 872
  3. 2019Conformal Prediction Under Covariate Shiftcited by 519
  4. 2021Adaptive Conformal Inference Under Distribution Shiftcited by 375
  5. 2023A Comprehensive Survey on Test-Time Adaptation under Distribution Shiftscited by 359
  6. 2022Conformal prediction beyond exchangeabilitycited by 355
  7. 2017Functional Map of the Worldcited by 338
  8. 2022Conformal Risk Controlcited by 299
  9. 2021Distribution-Free, Risk-Controlling Prediction Setscited by 288
  10. 2021Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Controlcited by 271
  11. 2023Prediction-Powered Inferencecited by 220
  12. 2020Estimating means of bounded random variables by bettingcited by 169
36 more
  1. 2022Adaptive Conformal Predictions for Time Seriescited by 161
  2. 2022Game-theoretic statistics and safe anytime-valid inferencecited by 155
  3. 2021Testing for Outliers with Conformal p-valuescited by 134
  4. 2021Localized Conformal Prediction: A Generalized Inference Framework for Conformal Predictioncited by 114
  5. 2022Is Out-of-Distribution Detection Learnable?cited by 109
  6. 2019Safe Testingcited by 101
  7. 2021Distribution-free uncertainty quantification for classification under label shiftcited by 101
  8. 2022Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shiftcited by 94
  9. 2022Wild-Time: A Benchmark of in-the-Wild Distribution Shift over Timecited by 90
  10. 2020Distribution-free binary classification: prediction sets, confidence intervals and calibrationcited by 79
  11. 2024Hypothesis testing with e-valuescited by 74
  12. 2024Online conformal prediction with decaying step sizescited by 56
  13. 2021Tracking the risk of a deployed model and detecting harmful distribution shiftscited by 40
  14. 2022Achieving Risk Control in Online Learning Settingscited by 39
  15. 2021Nonparametric Two-Sample Testing by Bettingcited by 37
  16. 2019Testing randomnesscited by 34
  17. 2021Retrain or not retrain: Conformal test martingales for change-point detectioncited by 32
  18. 2022E-detectors: a nonparametric framework for sequential change detectioncited by 32
  19. 2024Conformal Validity Guarantees Exist for Any Data Distribution (and How to Find Them)cited by 26
  20. 2024Protected Test-Time Adaptation via Online Entropy Matching: A Betting Approachcited by 23
  21. 2023Sequential Predictive Two-Sample and Independence Testingcited by 21
  22. 2017Inductive Conformal Martingales for Change-Point Detectioncited by 19
  23. 2023(Almost) Provable Error Bounds Under Distribution Shift via Disagreement Discrepancycited by 16
  24. 2024Adaptive Learn-then-Test: Statistically Valid and Efficient Hyperparameter Selectioncited by 15
  25. 2024Active, anytime-valid risk controlling prediction setscited by 14
  26. 2023Online multiple testing with e-valuescited by 12
  27. 2023When accurate prediction models yield harmful self-fulfilling propheciescited by 12
  28. 2022E-Valuating Classifier Two-Sample Testscited by 10
  29. 2023Reducing sequential change detection to sequential estimationcited by 10
  30. 2023On the near-optimality of betting confidence sets for bounded meanscited by 10
  31. 2024Sequential Harmful Shift Detection Without Labelscited by 10
  32. 2023Testing exchangeability by pairwise bettingcited by 9
  33. 2023Deep anytime-valid hypothesis testingcited by 7
  34. 2024Taming False Positives in Out-of-Distribution Detection with Human Feedbackcited by 7
  35. 2022Online Distribution Shift Detection via Recency Predictioncited by 6
  36. 2025Multiple testing in multi-stream sequential change detectioncited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 2; 2026: 8.

10 papers.

  1. 2025E-valuator: Reliable Agent Verifiers with Sequential Hypothesis Testingcited by 17
  2. 2025Monitoring Risks in Test-Time Adaptationcited by 11
  3. 2026Testing For Distribution Shifts with Conditional Conformal Test Martingalescited by 4
  4. 2026Prediction-Powered Risk Monitoring of Deployed Models for Detecting Harmful Distribution Shiftscited by 3
  5. 2026Certified Adaptive Refresh: Anytime-Valid Monitoring for Federated Conformal RAGcited by 2
  6. 2026Detecting Object Tracking Failure via Sequential Hypothesis Testingcited by 0
  7. 2026Incentive Aware AI Regulations: A Credal Characterisationcited by 0
  8. 2026Online Safety Monitoring for LLMscited by 0
  9. 2026When Can Old Evaluations Certify a New Model? Label-Efficient Release Decisions under Evaluator Driftcited by 0
  10. 2026Runtime Monitoring of Supermartingale Certificatescited by 0