mlsanityData to 14 Oct 2026

Test-Time Adaptation Induces Stronger Accuracy and Agreement-on-the-Line

arXiv 2310.04941 ↗First posted 7 Oct 2023cs.LG cs.AIcited by 11

Not on the map. Why?

Abstract (arXiv)

Recently, Miller et al. (2021) and Baek et al. (2022) empirically demonstrated strong linear correlations between in-distribution (ID) versus out-of-distribution (OOD) accuracy and agreement. These trends, coined accuracy-on-the-line (ACL) and agreement-on-the-line (AGL), enable OOD model selection and performance estimation without labeled data. However, these phenomena also break for certain shifts, such as CIFAR10-C Gaussian Noise, posing a critical bottleneck. In this paper, we make a key finding that recent test-time adaptation (TTA) methods not only improve OOD performance, but drastically strengthen the ACL and AGL trends in models, even in shifts where models showed very weak correlations before. To analyze this, we revisit the theoretical conditions from Miller et al. (2021) that outline the types of distribution shifts needed for perfect ACL in linear models. Surprisingly, these conditions are satisfied after applying TTA to deep models in the penultimate feature embedding space. In particular, TTA causes the data distribution to collapse complex shifts into those can be expressed by a singular scaling variable in the feature space. Our results show that by combining TTA with AGL-based estimation methods, we can estimate the OOD performance of models with high precision for a broader set of distribution shifts. This lends us a simple system for selecting the best hyperparameters and adaptation strategy without any OOD labeled data.

Key facts. Cited by 11 papers in the map (data through 14 Oct 2026). Builds on 52 of its 61 references that are in the map. First posted on arXiv 7 Oct 2023.

Builds on

52 of 61 references.

References · 52

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  4. 2021Swin Transformer: Hierarchical Vision Transformer using Shifted Windowscited by 12,139
  5. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 7,005
  6. 2016Densely Connected Convolutional Networkscited by 6,720
  7. 2016Layer Normalizationcited by 5,669
  8. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,795
  9. 2017Squeeze-and-Excitation Networkscited by 4,789
  10. 2020Training data-efficient image transformers & distillation through attentioncited by 4,586
  11. 2014Going Deeper with Convolutionscited by 4,241
  12. 2016Wide Residual Networkscited by 3,070
38 more
  1. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,877
  2. 2016Aggregated Residual Transformations for Deep Neural Networkscited by 2,546
  3. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,385
  4. 2016Identity Mappings in Deep Residual Networkscited by 1,965
  5. 2019Invariant Risk Minimizationcited by 1,854
  6. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,850
  7. 2019Do ImageNet Classifiers Generalize to ImageNet?cited by 1,515
  8. 2020Sharpness-Aware Minimization for Efficiently Improving Generalizationcited by 1,417
  9. 2020In Search of Lost Domain Generalizationcited by 1,054
  10. 2020Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptationcited by 848
  11. 2022Continual Test-Time Domain Adaptationcited by 525
  12. 2017Progressive Neural Architecture Searchcited by 487
  13. 2022Efficient Test-Time Model Adaptation without Forgettingcited by 469
  14. 2017Deep Layer Aggregationcited by 436
  15. 2020Improving robustness against common corruptions by covariate shift adaptationcited by 431
  16. 2022DeiT III: Revenge of the ViTcited by 419
  17. 2023Towards Stable Test-Time Adaptation in Dynamic Wild Worldcited by 415
  18. 2021MEMO: Test Time Robustness via Adaptation and Augmentationcited by 377
  19. 2021Accuracy on the Line: On the Strong Correlation Between Out-of-Distribution and In-Distribution Generalizationcited by 249
  20. 2022Test-Time Training with Masked Autoencoderscited by 212
  21. 2018Do CIFAR-10 Classifiers Generalize to CIFAR-10?cited by 200
  22. 2022Parameter-free Online Test-time Adaptationcited by 189
  23. 2022Leveraging Unlabeled Data to Predict Out-of-Distribution Performancecited by 145
  24. 2021Assessing Generalization of SGD via Disagreementcited by 117
  25. 2021Predicting with Confidence on Unseen Distributionscited by 116
  26. 2022Test-Time Adaptation via Conjugate Pseudo-labelscited by 98
  27. 2022Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shiftcited by 94
  28. 2021Extending the WILDS Benchmark for Unsupervised Adaptationcited by 86
  29. 2017Minimal-Entropy Correlation Alignment for Unsupervised Deep Domain Adaptationcited by 71
  30. 2023On Pitfalls of Test-Time Adaptationcited by 64
  31. 2022PolyLoss: A Polynomial Expansion Perspective of Classification Loss Functionscited by 63
  32. 2022Assaying Out-Of-Distribution Generalization in Transfer Learningcited by 53
  33. 2021Tune it the Right Way: Unsupervised Validation of Domain Adaptation via Soft Neighborhood Densitycited by 44
  34. 2022ID and OOD Performance Are Sometimes Inversely Correlated on Real-world Datasetscited by 44
  35. 2021If your data distribution shifts, use self-learningcited by 36
  36. 2020Why do classifier accuracies show linear trends under distribution shift?cited by 21
  37. 2022Three New Validators and a Large-Scale Benchmark Ranking for Unsupervised Domain Adaptationcited by 13
  38. 2023Demystifying Disagreement-on-the-Line in High Dimensionscited by 12

50 most cited of 52 All 52 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 3; 2025: 5; 2026: 3.

11 papers.

  1. 2024The Entropy Enigma: Success and Failure of Entropy Minimizationcited by 24
  2. 2024TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Modelscited by 23
  3. 2025Monitoring Risks in Test-Time Adaptationcited by 11
  4. 2024Accuracy on the wrong line: On the pitfalls of noisy data for out-of-distribution generalisationcited by 9
  5. 2025Less Data Less Tokens: Multilingual Unification Learning for Efficient Test-Time Reasoning in LLMscited by 3
  6. 2025Focus on Likely Classes for Test-Time Predictioncited by 1
  7. 2026When is Test-Time Adaptation Identifiable From Unlabeled Evidence?cited by 1
  8. 2025Higher-Order Asymptotics of Test-Time Adaptation for Batch Normalization Statisticscited by 0
  9. 2025Data Curation Matters: Model Collapse and Spurious Shift Performance Prediction from Training on Uncurated Text Embeddingscited by 0
  10. 2026Autonomous Drift Learning in Data Streams: A Unified Perspectivecited by 0
  11. 2026Entropy Minimization without Model Collapse: Mitigating Prediction Bias in Medical Imagingcited by 0