mlsanityData to 7 Oct 2026

Open-Set Domain Adaptation Under Background Distribution Shift: Challenges and A Provably Efficient Solution

arXiv 2512.01152 ↗First posted 1 Dec 2025cs.LG cs.AI cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

As we deploy machine learning systems in the real world, a core challenge is to maintain a model that is performant even as the data shifts. Such shifts can take many forms: new classes may emerge that were absent during training, a problem known as open-set recognition, and the distribution of known categories may change. Guarantees on open-set recognition are mostly derived under the assumption that the distribution of known classes, which we call the background distribution, is fixed. In this paper we develop CoLOR, a method that is guaranteed to solve open-set recognition even in the challenging case where the background distribution shifts. We prove that the method works under benign assumptions that the novel class is separable from the non-novel classes, and provide theoretical guarantees that it outperforms a representative baseline in a simplified overparameterized setting. We develop techniques to make CoLOR scalable and robust, and perform comprehensive empirical evaluations on image and text data. The results show that CoLOR significantly outperforms existing open-set recognition methods under background shift. Moreover, we provide new insights into how factors such as the size of the novel class influences performance, an aspect that has not been extensively explored in prior work.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 48 of its 74 references that are in the map. First posted on arXiv 1 Dec 2025.

Builds on

48 of 74 references.

References · 48

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  3. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
  4. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  5. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,362
  6. 2016Concrete Problems in AI Safetycited by 1,582
  7. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  8. 2017Deep Hashing Network for Unsupervised Domain Adaptationcited by 1,181
  9. 2018Reconciling modern machine learning practice and the bias-variance trade-offcited by 919
  10. 2020Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptationcited by 844
  11. 2017The Implicit Bias of Gradient Descent on Separable Datacited by 649
  12. 2015Towards Open Set Deep Networkscited by 533
36 more
  1. 2017VisDA: The Visual Domain Adaptation Challengecited by 432
  2. 2020A Unifying Review of Deep and Shallow Anomaly Detectioncited by 423
  3. 2021Open-Set Recognition: a Good Closed-Set Classifier is All You Need?cited by 351
  4. 2020An Investigation of Why Overparameterization Exacerbates Spurious Correlationscited by 310
  5. 2018Learning from positive and unlabeled data: a surveycited by 218
  6. 2017Positive-Unlabeled Learning with Non-Negative Risk Estimatorcited by 217
  7. 2015The Benefit of Multitask Representation Learningcited by 194
  8. 2018Reducing Network Agnostophobiacited by 165
  9. 2017Generative OpenMax for Multi-Class Open Set Classificationcited by 157
  10. 2021Adversarial Reciprocal Points Learning for Open Set Recognitioncited by 153
  11. 2021Zero-Shot Out-of-Distribution Detection Based on the Pre-trained Model CLIPcited by 151
  12. 2018Open Set Domain Adaptation by Backpropagationcited by 150
  13. 2020Understanding the Failure Modes of Out-of-Distribution Generalizationcited by 146
  14. 2020Can Autonomous Vehicles Identify, Recover From, and Adapt to Distribution Shifts?cited by 139
  15. 2021OpenGAN: Open-Set Recognition via Open Data Generationcited by 135
  16. 2020A survey on domain adaptation theory: learning bounds and theoretical guaranteescited by 109
  17. 2018Optimization with Non-Differentiable Constraints with Applications to Fairness, Recall, Churn, and Other Goalscited by 103
  18. 2020Classification vs regression in overparameterized regimes: Does the loss function matter?cited by 97
  19. 2019Identifying Unknown Instances for Autonomous Drivingcited by 73
  20. 2021Mixture Proportion Estimation and PU Learning: A Modern Approachcited by 52
  21. 2021Constrained Learning with Non-Convex Lossescited by 42
  22. 2021Modeling Discriminative Representations for Out-of-Domain Detection with Supervised Contrastive Learningcited by 42
  23. 2018Open-world Learning and Application to Product Classificationcited by 39
  24. 2018Beyond the Selected Completely At Random Assumption for Learning from Positive and Unlabeled Datacited by 36
  25. 2018Open Category Detection with PAC Guaranteescited by 35
  26. 2023Anomaly Detection under Distribution Shiftcited by 35
  27. 2022Domain Adaptation under Open Set Label Shiftcited by 32
  28. 2023Don't blame Dataset Shift! Shortcut Learning due to Gradients and Cross Entropycited by 30
  29. 2023Universal Domain Adaptation via Compressive Attention Matchingcited by 15
  30. 2024Open-Set Domain Adaptation for Semantic Segmentationcited by 13
  31. 2020A Post-processing Method for Detecting Unknown Intent of Dialogue System via Pre-trained Deep Neural Network Classifiercited by 11
  32. 2023Model-free Test Time Adaptation for Out-Of-Distribution Detectioncited by 9
  33. 2022Learning Classifiers of Prototypes and Reciprocal Points for Universal Domain Adaptationcited by 8
  34. 2022Malign Overfitting: Interpolation Can Provably Preclude Invariancecited by 6
  35. 2024Cross-domain Open-world Discoverycited by 6
  36. 2024SKADA-Bench: Benchmarking Unsupervised Domain Adaptation Methods with Realistic Validation On Diverse Modalitiescited by 2

Built on by

Not cited yet.