mlsanityData to 7 Oct 2026

Towards Stable Test-Time Adaptation in Dynamic Wild World

arXiv 2302.12400 ↗First posted 24 Feb 2023cs.LG cs.CVcited by 409

Not on the map. Why?

Cited by 39 in Multimodal LLMs / vision-language models, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Test-time adaptation (TTA) has shown to be effective at tackling distribution shifts between training and testing data by adapting a given model on test samples. However, the online model updating of TTA may be unstable and this is often a key obstacle preventing existing TTA methods from being deployed in the real world. Specifically, TTA may fail to improve or even harm the model performance when test data have: 1) mixed distribution shifts, 2) small batch sizes, and 3) online imbalanced label distribution shifts, which are quite common in practice. In this paper, we investigate the unstable reasons and find that the batch norm layer is a crucial factor hindering TTA stability. Conversely, TTA can perform more stably with batch-agnostic norm layers, ie, group or layer norm. However, we observe that TTA with group and layer norms does not always succeed and still suffers many failure cases. By digging into the failure cases, we find that certain noisy test samples with large gradients may disturb the model adaption and result in collapsed trivial solutions, ie, assigning the same class label for all samples. To address the above collapse issue, we propose a sharpness-aware and reliable entropy minimization method, called SAR, for further stabilizing TTA from two aspects: 1) remove partial noisy samples with large gradients, 2) encourage model weights to go to a flat minimum so that the model is robust to the remaining noisy samples. Promising results demonstrate that SAR performs more stably over prior methods and is computationally efficient under the above wild test scenarios.

Key facts. Cited by 409 papers in the map (data through 7 Oct 2026). Builds on 56 of its 76 references that are in the map. First posted on arXiv 24 Feb 2023.

Builds on

56 of 76 references.

References · 56

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 11,979
  3. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  4. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 6,979
  5. 2016Layer Normalizationcited by 5,579
  6. 2022A ConvNet for the 2020scited by 3,545
  7. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  8. 2017Non-local Neural Networkscited by 2,290
  9. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,842
  10. 2018Unsupervised Representation Learning by Predicting Image Rotationscited by 1,586
  11. 2018Group Normalizationcited by 1,514
  12. 2020Sharpness-Aware Minimization for Efficiently Improving Generalizationcited by 1,404
38 more
  1. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  2. 2017Visualizing the Loss Landscape of Neural Netscited by 1,126
  3. 2017StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translationcited by 1,036
  4. 2019AugMix: A Simple Data Processing Method to Improve Robustness and Uncertaintycited by 897
  5. 2020Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptationcited by 844
  6. 2017Learning to Generalize: Meta-Learning for Domain Generalizationcited by 762
  7. 2017Maximum Classifier Discrepancy for Unsupervised Domain Adaptationcited by 682
  8. 2022Continual Test-Time Domain Adaptationcited by 520
  9. 2022Efficient Test-Time Model Adaptation without Forgettingcited by 467
  10. 2020Improving robustness against common corruptions by covariate shift adaptationcited by 427
  11. 2019Domain Generalization via Model-Agnostic Learning of Semantic Featurescited by 405
  12. 2021MEMO: Test Time Robustness via Adaptation and Augmentationcited by 376
  13. 2017On Kernelized Multi-armed Banditscited by 348
  14. 2021ASAM: Adaptive Sharpness-Aware Minimization for Scale-Invariant Learning of Deep Neural Networkscited by 293
  15. 2018Generalizing Across Domains via Cross-Gradient Trainingcited by 268
  16. 2021When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentationscited by 259
  17. 2022Contrastive Test-Time Adaptationcited by 258
  18. 2019Fast AutoAugmentcited by 243
  19. 2020Evaluating Prediction-Time Batch Normalization for Robustness under Covariate Shiftcited by 235
  20. 2020Universal Source-Free Domain Adaptationcited by 219
  21. 2018Online Learning: A Comprehensive Surveycited by 203
  22. 2022Improving Out-of-Distribution Robustness via Selective Augmentationcited by 202
  23. 2018Do CIFAR-10 Classifiers Generalize to CIFAR-10?cited by 197
  24. 2018Multi-Adversarial Domain Adaptationcited by 195
  25. 2021Efficient Sharpness-aware Minimization for Improved Training of Neural Networkscited by 137
  26. 2021Source-free Domain Adaptation via Avatar Prototype Generation and Adaptationcited by 111
  27. 2023TTN: A Domain-Shift Aware Batch Normalization in Test-Time Adaptationcited by 101
  28. 2020Regularizing Neural Networks via Adversarial Model Perturbationcited by 78
  29. 2023DELTA: degradation-free fully test-time adaptationcited by 72
  30. 2020On Feature Normalization and Data Augmentationcited by 71
  31. 2021Test-Time Adaptation to Distribution Shift by Confidence Maximization and Input Transformationcited by 67
  32. 2022Improving Test-Time Adaptation via Shift-agnostic Weight Regularization and Nearest Source Prototypescited by 67
  33. 2021MT3: Meta Test-Time Training for Self-Supervised Test-Time Adaptioncited by 65
  34. 2020Collaborative Unsupervised Domain Adaptation for Medical Image Diagnosiscited by 60
  35. 2021SITA: Single Image Test-time Adaptationcited by 57
  36. 2021MixNorm: Test-Time Adaptation Through Online Normalization Estimationcited by 47
  37. 2021Rethinking "Batch" in BatchNormcited by 44
  38. 2022Prototype-Guided Continual Adaptation for Class-Incremental Unsupervised Domain Adaptationcited by 33

50 most cited of 56 All 56 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 56; 2024: 115; 2025: 110; 2026: 128.

409 papers.

Sort:

By area:

  1. 2023A Comprehensive Survey on Test-Time Adaptation under Distribution Shiftscited by 353
  2. 2023AdaMerging: Adaptive Model Merging for Multi-Task Learningcited by 248
  3. 2023Revisiting Class-Incremental Learning with Pre-Trained Models: Generalizability and Adaptivity are All You Needcited by 183
  4. 2023EcoTTA: Memory-Efficient Continual Test-time Adaptation via Self-distilled Regularizationcited by 123
  5. 2024Entropy is not Enough for Test-Time Adaptation: From the Perspective of Disentangled Factorscited by 106
  6. 2024Test-Time Model Adaptation with Only Forward Passescited by 74
  7. 2023A Comprehensive Survey of Forgetting in Deep Learning Beyond Continual Learningcited by 66
  8. 2023Universal Test-time Adaptation through Weight Ensembling, Diversity Weighting, and Prior Correctioncited by 65
  9. 2023On Pitfalls of Test-Time Adaptationcited by 63
  10. 2023SoTTA: Robust Test-Time Adaptation on Noisy Data Streamscited by 63
  11. 2023ViDA: Homeostatic Visual Domain Adapter for Continual Test Time Adaptationcited by 58
  12. 2024A Survey on All-in-One Image Restoration: Taxonomy, Evaluation and Future Trendscited by 58
38 more
  1. 2025Test-Time Learning for Large Language Modelscited by 46
  2. 2023In Search of Lost Online Test-time Adaptation: A Surveycited by 43
  3. 2025NavMorph: A Self-Evolving World Model for Vision-and-Language Navigation in Continuous Environmentscited by 41
  4. 2025A Survey of Test-Time Compute: From Intuitive Inference to Deliberate Reasoningcited by 40
  5. 2023Towards Open-Set Test-Time Adaptation Utilizing the Wisdom of Crowds in Entropy Minimizationcited by 39
  6. 2023Test-Time Adaptation with CLIP Reward for Zero-Shot Generalization in Vision-Language Modelscited by 37
  7. 2024Knowledge Composition using Task Vectors with Learned Anisotropic Scalingcited by 37
  8. 2024Frustratingly Easy Test-Time Adaptation of Vision-Language Modelscited by 36
  9. 2024Dual Prototype Evolving for Test-Time Generalization of Vision-Language Modelscited by 36
  10. 2023On the Robustness of Open-World Test-Time Training: Self-Training with Dynamic Prototype Expansioncited by 34
  11. 2023Towards Real-World Test-Time Adaptation: Tri-Net Self-Training with Balanced Normalizationcited by 34
  12. 2024DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptationcited by 33
  13. 2024Just Shift It: Test-Time Prototype Shifting for Zero-Shot Generalization with Vision-Language Modelscited by 31
  14. 2024COME: Test-time adaption by Conservatively Minimizing Entropycited by 31
  15. 2023Continual-MAE: Adaptive Distribution Masked Autoencoders for Continual Test-Time Adaptationcited by 30
  16. 2023Each Test Image Deserves A Specific Prompt: Continual Test-Time Adaptation for 2D Medical Image Segmentationcited by 29
  17. 2023Exploring Sparse Visual Prompt for Domain Adaptive Dense Predictioncited by 28
  18. 2023Fast-Slow Test-Time Adaptation for Online Vision-and-Language Navigationcited by 25
  19. 2023TEA: Test-time Energy Adaptationcited by 25
  20. 2024BECoTTA: Input-dependent Online Blending of Experts for Continual Test-time Adaptationcited by 25
  21. 2024Any-Shift Prompting for Generalization over Distributionscited by 24
  22. 2024The Entropy Enigma: Success and Failure of Entropy Minimizationcited by 24
  23. 2024Protected Test-Time Adaptation via Online Entropy Matching: A Betting Approachcited by 22
  24. 2023Label Shift Adapter for Test-Time Adaptation under Covariate and Label Shiftscited by 21
  25. 2025Bayesian Test-Time Adaptation for Vision-Language Modelscited by 21
  26. 2023Detecting Adversarial Data by Probing Multiple Perturbations Using Expected Perturbation Scorecited by 20
  27. 2024Uncertainty-Calibrated Test-Time Model Adaptation without Forgettingcited by 20
  28. 2024Proxy Denoising for Source-Free Domain Adaptationcited by 20
  29. 2023Efficient Test-Time Adaptation for Super-Resolution with Second-Order Degradation and Reconstructioncited by 18
  30. 2023Persistent Test-time Adaptation in Recurring Testing Scenarioscited by 18
  31. 2024Active Test-Time Adaptation: Theoretical Analyses and An Algorithmcited by 18
  32. 2023Benchmarking Test-Time Adaptation against Distribution Shifts in Image Classificationcited by 17
  33. 2023Test-Time Degradation Adaptation for Open-Set Image Restorationcited by 17
  34. 2024Unified Entropy Optimization for Open-Set Test-Time Adaptationcited by 17
  35. 2024OT-VP: Optimal Transport-guided Visual Prompting for Test-Time Adaptationcited by 17
  36. 2024BoostAdapter: Improving Vision-Language Test-Time Adaptation via Regional Bootstrappingcited by 17
  37. 2025Ranked Entropy Minimization for Continual Test-Time Adaptationcited by 17
  38. 2026AdaJEPA: An Adaptive Latent World Modelcited by 16

50 most cited of 409 All 409 (JSON)