mlsanityData to 13 Oct 2026

Ensuring Calibration Robustness in Split Conformal Prediction Under Adversarial Attacks

arXiv 2511.18562 ↗First posted 23 Nov 2025stat.ML cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Conformal prediction (CP) provides distribution-free, finite-sample coverage guarantees but critically relies on exchangeability, a condition often violated under distribution shift. We study the robustness of split conformal prediction under adversarial perturbations at test time, focusing on both coverage validity and the resulting prediction set size. Our theoretical analysis characterizes how the strength of adversarial perturbations during calibration affects coverage guarantees under adversarial test conditions. We further examine the impact of adversarial training at the model-training stage. Extensive experiments support our theory: (i) Prediction coverage varies monotonically with the calibration-time attack strength, enabling the use of nonzero calibration-time attack to predictably control coverage under adversarial tests; (ii) target coverage can hold over a range of test-time attacks: with a suitable calibration attack, coverage stays within any chosen tolerance band across a contiguous set of perturbation levels; and (iii) adversarial training at the training stage produces tighter prediction sets that retain high informativeness.

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Builds on 39 of its 62 references that are in the map. First posted on arXiv 23 Nov 2025.

Builds on

39 of 62 references.

References · 39

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,043
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,336
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,379
  4. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,511
  5. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,371
  6. 2017Evasion Attacks against Machine Learning at Test Timecited by 689
  7. 2016Distribution-Free Predictive Inference For Regressioncited by 660
  8. 2019Adversarial Training for Free!cited by 641
  9. 2018Bag of Tricks for Image Classification with Convolutional Neural Networkscited by 430
  10. 2019Unlabeled Data Improves Adversarial Robustnesscited by 416
  11. 2020Classification with Valid and Adaptive Coveragecited by 393
  12. 2022Conformal prediction beyond exchangeabilitycited by 351
27 more
  1. 2019Adversarial Examples Improve Image Recognitioncited by 284
  2. 2021Recent Advances in Adversarial Training for Adversarial Robustnesscited by 278
  3. 2021When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentationscited by 259
  4. 2018Generating Adversarial Examples with Adversarial Networkscited by 258
  5. 2020Understanding and Improving Fast Adversarial Trainingcited by 213
  6. 2015Learning with Symmetric Label Noise: The Importance of Being Unhingedcited by 120
  7. 2017Delving into adversarial attacks on deep policiescited by 104
  8. 2021On the Adversarial Robustness of Vision Transformerscited by 104
  9. 2018Universal Adversarial Trainingcited by 86
  10. 2020Precise Tradeoffs in Adversarial Training for Linear Regressioncited by 77
  11. 2020Opportunities and Challenges in Deep Learning Adversarial Robustness: A Surveycited by 71
  12. 2022Practical Adversarial Multivalid Conformal Predictioncited by 65
  13. 2024Conformal Alignment: Knowing When to Trust Foundation Models with Guaranteescited by 52
  14. 2021Reveal of Vision Transformers Robustness against Adversarial Attackscited by 43
  15. 2025Decision Theoretic Foundations for Conformal Prediction: Optimal Uncertainty Quantification for Risk-Averse Agentscited by 39
  16. 2024Provably Robust Conformal Prediction with Improved Efficiencycited by 27
  17. 2023Probabilistically robust conformal predictioncited by 25
  18. 2020On the Generalization Properties of Adversarial Trainingcited by 23
  19. 2022Robustness to Label Noise Depends on the Shape of the Noise Distribution in Feature Spacecited by 15
  20. 2024Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Datacited by 12
  21. 2024COLEP: Certifiably Robust Learning-Reasoning Conformal Prediction via Probabilistic Circuitscited by 9
  22. 2025Conformal Prediction and Human Decision Makingcited by 9
  23. 2025Enhancing Adversarial Robustness with Conformal Prediction: A Framework for Guaranteed Model Reliabilitycited by 6
  24. 2024Provably Reliable Conformal Prediction Sets in the Presence of Data Poisoningcited by 4
  25. 2025Efficient Robust Conformal Prediction via Lipschitz-Bounded Networkscited by 4
  26. 2019The Attack Generator: A Systematic Approach Towards Constructing Adversarial Attackscited by 3
  27. 2023Adversarial Training with Generated Data in High-Dimensional Regression: An Asymptotic Studycited by 1

Built on by

Not cited yet.