Pith. sign in

REVIEW 2 major objections 6 minor 56 references

Label Distribution Shift-Aware Prediction Refinement for Test-Time Adaptation

T0 review · 2 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A two-layer module trained on imbalanced batches rescues test-time adaptation when class labels shift.

desk verdict Solid plug-in for label-shift-aware TTA, but the no-degradation claim is overbroad and the train/test distribution gap for the refinement module is unverified. read the letter →

arxiv 2411.15204 v2 pith:JZZOJUVF submitted 2024-11-20 cs.LG cs.AIcs.CV

Minguk Jang , Hye Won Chung This is my paper · ORCID
classification cs.LGcs.AIcs.CV
keywords test-timeadaptationlabeldistributionshiftpredictionrefinementbatchnormalizationDirichletsamplingclassimbalancepseudo-labelaccuracyaffinelogitcorrection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that the performance collapse of batch-normalization-based test-time adaptation (TTA) methods under test-time label distribution shift is systematic and correctable. It identifies consistent class-wise confusion patterns in BN-adapted classifiers under long-tailed test label distributions, and introduces DART, a small prediction refinement module trained during an intermediate time on Dirichlet-sampled batches. DART detects the direction and severity of the label shift from the average pseudo-label distribution and a confidence measure called prediction deviation, then applies an affine correction to logits. Plugging DART into BNAdapt and six other TTA methods recovers most of the accuracy lost under imbalance ratios up to 100 while leaving balanced performance unchanged, including gains of 5.7% and 18.1% for BNAdapt at ratios 10 and 100.

What carries the argument

The central object is the prediction refinement module $g_\phi$, a 2-layer MLP that maps a batch's averaged pseudo-label distribution $\bar{p}_B$ and prediction deviation $d_B$ to a square matrix $W_B \in \mathbb{R}^{K \times K}$ and a bias vector $b_B \in \mathbb{R}^K$. At test time the classifier's logits are refined as $\mathrm{softmax}(\bar{f}_\theta(x)W_B + b_B)$; the module is trained during an intermediate time using cross-entropy on Dirichlet-sampled imbalanced batches regularized to output identity/zero on balanced batches. This affine correction is the mechanism that reverses the class-wise confusion patterns.

What would settle it

Train DART using only a subset of corruption types (e.g., noise-type corruptions) and evaluate on held-out corruption types (e.g., blur or weather types) under the same label imbalance; if accuracy falls back to the BNAdapt level, the cross-corruption consistency claim fails. Conversely, one could compute the least-squares optimal affine correction for each test batch using ground-truth labels and compare it with the module's output; a large divergence at moderate imbalance would invalidate the learned correction.

Watch

Extended reading notes

Core claim

The central claim is that a classifier's confusion pattern under combined covariate and label shift is an affine function of the shifted label distribution, consistent enough across corruption types to be learned from simulated Dirichlet-sampled batches of the training data. The paper argues that a mapping from (average pseudo-label distribution, prediction deviation) to a logit transform $(W_B, b_B)$ can be learned by a 2-layer MLP, and that applying this transform reverses the class-wise confusion caused by the label shift. In experiments, this recovers virtually all accuracy lost by BNAdapt under long-tailed test distributions (e.g., CIFAR-10C-LT at imbalance ratio 100 improves from 67.0% to 85.1%) with no degradation when the label distribution is balanced.

Load-bearing premise

The load-bearing premise is that the class-wise confusion pattern of a BN-adapted classifier under label shift is consistent enough across corruption types, and similar between Dirichlet-sampled clean training batches and real test batches, that a 2-layer MLP trained only on those simulated batches can output a correct affine correction at test time.

Editorial extensions

If this is right

  • BNAdapt+DART lifts CIFAR-10C-LT accuracy at imbalance ratio 100 from 67.0% to 85.1%, close to the balanced-case level of 85.2%.
  • The plug-in improves every tested TTA baseline (TENT, PL, NOTE, LAME, DELTA, ODS, SAR) under label distribution shift, with the largest gains where pseudo-label accuracy is the bottleneck.
  • DART requires only soft pseudo-labels at test time and adds negligible test-time compute, since the refinement module is a 2-layer MLP that stays fixed.
  • On large-scale benchmarks, DART-split (which separates shift detection from transformation generation) keeps accuracy above the no-adaptation baseline even at extreme imbalance ratios such as IR 50000 on CIFAR-100C.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The consistency claim suggests that confusion patterns are dominated by BN-statistics bias rather than corruption specifics, implying that a single learned correction should transfer to unseen corruption types that share the same label distribution.
  • The module effectively learns a label-shift-dependent logit adjustment without test labels, which connects to long-tail logit-adjustment methods and could be combined with them for online settings.
  • Prediction deviation appears to act as a monotone severity signal (as shown in the paper's Figure 4), so it could be reused as an adaptive weighting or early-stopping criterion in other TTA pipelines.
  • A testable extension is to train DART on only a few corruption types and evaluate on held-out corruptions; the paper's consistency premise predicts the correction should still transfer.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper proposes DART, a prediction refinement module trained during an 'intermediate time' on labeled training batches whose class distributions are sampled from a Dirichlet distribution. The module gφ maps two batch-level statistics—the average pseudo-label distribution p̄_B and a prediction deviation d_B—to an affine transformation (W_B,b_B) of the classifier logits. At test time, this transform is applied to the predictions of existing TTA methods (BNAdapt, TENT, PL, NOTE, LAME, DELTA, ODS, SAR) to correct the class-wise confusion caused by combined covariate and label shift. The authors report large gains on CIFAR-10C-LT (e.g., BNAdapt+DART 85.1 vs 67.0 at ρ=100), consistent improvements on CIFAR-10C-imb, CIFAR-100C-imb, ImageNet-C-imb, PACS, OfficeHome, and DomainNet-126, and introduce a DART-split variant for large label spaces.

Significance. If the reported results hold, DART is a practically valuable plug-in for TTA under label distribution shift, with gains of 5–18 percentage points on CIFAR-10C-LT and consistent improvements across many baselines and benchmarks. The paper's strengths include: extensive experiments with four seeds, per-corruption tables in the appendix, ablations of the module's inputs/outputs, sensitivity checks for α and architecture size, and a training protocol that uses no test labels. The main risks are (i) the blanket claim of 'no performance degradation' at ρ=1, which is contradicted by the paper's own per-corruption results (Table 19), and (ii) the unverified extrapolation of gφ outside the support of the intermediate-time (p̄,d) distribution, which is the load-bearing premise for transfer to unseen corruptions and extreme imbalance.

major comments (2)
  1. [Abstract; §4.1; Table 19] The claim that DART causes no performance degradation when there is no label distribution shift is contradicted by the per-corruption results in Table 19. At ρ=1, BNAdapt+DART drops from 69.9±0.1 to 68.1±0.7 on impulse_noise, TENT+DART drops from 74.0±0.4 to 71.8±1.5 on impulse_noise, PL+DART drops from 74.1±0.5 to 72.5±0.8, and several NOTE+DART entries drop by 0.5–2.9 points (e.g., impulse_noise 65.9→63.0, defocus_blur 88.6→88.0). The aggregate ρ=1 averages in Table 1 are indeed unchanged (85.2 vs 85.2 for BNAdapt), but the Abstract and Section 1 state a stronger, unconditional claim ('without any performance degradation'). The authors should either restrict the claim to average accuracy or explicitly discuss and explain the per-corruption drops, since a user applying DART to a single corruption (e.g., impulse_noise) would see a measurable accuracy loss.
  2. [§3, Eq. (4); Appendix A.3; Table 4] The central mechanism assumes that the affine correction learned by gφ on Dirichlet-sampled clean training batches transfers to test batches under combined covariate and label shift, but the paper never checks whether the test-time input statistics (p̄_B,d_B) lie inside the support of the intermediate-time training distribution. For CIFAR-10, δ=10 yields near-uniform class distributions (Appendix A.3), whereas test batches at ρ=100 have strongly skewed p̄_B and markedly lower d_B; corrupted inputs also shift d_B below the clean-training range. Table 4 shows that the method is acutely sensitive to the intermediate sampling distribution (the Unif&LT variant collapses from 82.4 to 28.7 at IR5000), so coverage of the (p̄,d) space is load-bearing. I request a direct analysis: plot or quantify the joint distribution of (p̄_B,d_B) for intermediate-time batches and for test batches per corruption and ρ/IR, and either demonstrate that test inputs fall within the training support or provide an explicit extrapolation analysis, including failure cases such as the impulse_noise degradation noted above.
minor comments (6)
  1. [Appendix D.2, Eqs. (28) and (31)] The derivation of W* assumes logit invariance to the class distribution in Eq. (28) and approximates (x−pμ)≈(x−qμ) in Eq. (31) without stating conditions under which these hold; as written, the 'theoretical meaning' of the affine correction is not established. Please either prove the invariance under stated assumptions or present Section D.2 as a heuristic interpretation rather than a derivation.
  2. [Appendix A.3, A.4, B.2] Several hyperparameters are tuned per benchmark (δ, Ndir, softmax temperature scaling, DART-split threshold, and α=10 for PACS/OfficeHome). Tables 10 and 14 provide sensitivity for α, intermediate batch size, and hidden dimension, but not for δ or Ndir. Please report sensitivity to these choices or justify the per-dataset values.
  3. [Section 2; Figures 1, 9–11] The claim that class-wise confusion patterns are 'consistent across different types of corruption' is supported only visually. Please provide a quantitative similarity measure (e.g., mean absolute difference or correlation of off-diagonal entries between corruption types) to strengthen this empirical claim.
  4. [Table 20] The 'ODS+DART (ours)' row header appears to contain a duplicated '(ours)' fragment, which should be cleaned up.
  5. [Abstract; Section 4.1] The '5–18% gains' range in the Abstract applies to BNAdapt; for other baselines the gains can be smaller (e.g., TENT+DART at ρ=10 is about 3.8 pp). Consider clarifying that the stated range refers to the BNAdapt case.
  6. [Section 4.1] The sentence 'maintaining the original performance for ρ=1' should be qualified to 'maintaining the average performance' in view of the per-corruption drops in Table 19.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DART's refinement module is trained on labeled clean batches and tested on unseen test batches, so the central claim is not equivalent to its inputs.

full rationale

The central training/test separation is non-circular. In Equations (1)-(3) and Algorithm 1, g_phi is fit on labeled Dirichlet-sampled batches from the clean training data before test time, using cross-entropy against ground-truth labels of those intermediate batches. At test time, Equation (4) computes only the averaged pseudo-label distribution and prediction deviation from unlabeled test data, and g_phi is frozen; no test labels are used to fit the correction. The theoretical analysis in Section D.1 derives confusion-pattern properties from a Gaussian toy model and does not assume the target result. The paper's own ablation in Table 4 shows the method is sensitive to the intermediate sampling distribution, which confirms that the test-time correction is an extrapolation rather than a tautology. The only self-citation by the authors, reference [11], appears as background for nearest-neighbor TTA methods in Appendix C.2 and is not load-bearing. The main limitation is a coverage/generalization risk for out-of-support test statistics, not a logical circularity.

Assumptions & free parameters 8 free parameters · 7 assumptions · 3 invented entities

The core method is supervised learning of a correction function; the main unfunded assumptions are transferability of confusion patterns and the sufficiency of the two batch statistics. The theoretical derivation in Appendix D adds two unjustified algebraic steps. No new physical entities are postulated, but the learned module and metric are new constructs with empirical support.

free parameters (8)
  • Regularization weight α = 0.1 for most benchmarks; 10 for PACS and OfficeHome
    Balances imbalanced vs balanced loss in Eq. (3); hand-tuned per benchmark (Appendix A.3).
  • Dirichlet concentration δ = 10 for CIFAR-10/PACS/OfficeHome; 0.1 for CIFAR-100; 0.001 for ImageNet-C
    Controls diversity of intermediate class distributions; set per benchmark (Appendix A.3).
  • Number of Dirichlet chunks Ndir = 250 for CIFAR-10/PACS/OfficeHome; 1000 for CIFAR-100C; 2000 for ImageNet-C
    Number of Dirichlet-sampled batches used in intermediate training (Appendix A.3).
  • Softmax temperature scaling factor = 2 for OfficeHome; 1.1 for CIFAR-100; 1.1 for CIFAR-10C in DART-split
    Applied to test logits to control confidence; set per benchmark (Appendix A.4, B.2).
  • Intermediate batch size = 64 for CIFAR/PACS/OfficeHome; 32 for ImageNet
    Batch size used for intermediate-time training (Appendix A.3).
  • Hidden dimension of gφ = 1000
    MLP hidden size; chosen without full sensitivity analysis across all benchmarks, though Table 14 shows some robustness.
  • Intermediate training epochs = 50 for CIFAR-10; 100 for CIFAR-100; 200 for ImageNet
    Training epochs for gφ; Table 15 shows 10 epochs suffices on CIFAR-10C-LT.
  • DART-split severity threshold = 0.5
    Gating threshold on sigmoid severity score s_B for DART-split (Algorithm 3).
assumptions (7)
  • domain assumption Confusion patterns of a BN-adapted classifier under label shift are consistent across corruption types and reflect class relationships and shift magnitude and direction.
    Motivates transferring gφ trained on clean training data to corrupted test data; stated in Section 2 and Fig. 1.
  • domain assumption The averaged pseudo-label distribution and prediction deviation are sufficient inputs to determine the affine correction W, b.
    gφ is a 2-layer MLP mapping these two statistics to a K by K matrix and bias; no proof of sufficiency, especially for large K.
  • ad hoc to paper Logits are robust to changes in class distribution: Eq. (28) sets l(x) = (x − pμ)μ^T(I − 1p)^T equal to (x − rμ)μ^T(I − 1r)^T for arbitrary r.
    Used to derive W* in Appendix D.2; does not hold for general class centroids μ, since the equality implies (I − 1p) = (I − 1r) or degenerate μ.
  • ad hoc to paper Approximation (x − pμ) ≈ (x − qμ) in Eq. (31).
    Needed to derive the closed-form W*; the difference is (q − p)μ, not justified to be small.
  • domain assumption The sum of logits is approximately constant for training and test data in Eq. (7).
    Used for the softmax temperature scaling T formula in Appendix B.1; stated as an assumption.
  • domain assumption Mean centering approximates batch normalization in BNAdapt.
    Toy model in Section D.1 treats BN as mean centering only, ignoring variance normalization and affine parameters.
  • domain assumption Labeled training data is available during the intermediate time.
    DART requires the training dataset between training and test; authors acknowledge this in Section 6.1 and also test a condensed-data variant.
invented entities (3)
  • Prediction refinement module gφ independent evidence
    purpose: Learns affine logit correction (W, b) from batch pseudo-label statistics
    Trained on labeled training batches, applied to unseen test benchmarks; held-out results in Table 1 support its effectiveness.
  • Prediction deviation metric d_B independent evidence
    purpose: Measures average per-sample divergence of softmax outputs from uniform, used to detect severity of label shift
    Figure 4 shows a monotonic relationship between d_B and test accuracy across imbalance ratios; not a fitted constant.
  • DART-split severity score s_B independent evidence
    purpose: Gating scalar deciding whether to apply affine refinement per test batch
    Trained by binary classification on intermediate batches; threshold fixed at 0.5.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Label Distribution Shift-Aware Prediction Refinement for Test-Time Adaptation." pith.science (2026). https://pith.science/paper/JZZOJUVF

@misc{pith2026241115204,
  author       = {Pith},
  title        = {Pith review of: Label Distribution Shift-Aware Prediction Refinement for Test-Time Adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JZZOJUVF}},
  note         = {Machine review of arXiv:2411.15204}
}
read the original abstract

Test-time adaptation (TTA) is an effective approach to mitigate performance degradation of trained models when encountering input distribution shifts at test time. However, existing TTA methods often suffer significant performance drops when facing additional class distribution shifts. We first analyze TTA methods under label distribution shifts and identify the presence of class-wise confusion patterns commonly observed across different covariate shifts. Based on this observation, we introduce label Distribution shift-Aware prediction Refinement for Test-time adaptation (DART), a novel TTA method that refines the predictions by focusing on class-wise confusion patterns. DART trains a prediction refinement module during an intermediate time by exposing it to several batches with diverse class distributions using the training dataset. This module is then used during test time to detect and correct class distribution shifts, significantly improving pseudo-label accuracy for test data. Our method exhibits 5-18% gains in accuracy under label distribution shifts on CIFAR-10C, without any performance degradation when there is no label distribution shift. Extensive experiments on CIFAR, PACS, OfficeHome, and ImageNet benchmarks demonstrate DART's ability to correct inaccurate predictions caused by test-time distribution shifts. This improvement leads to enhanced performance in existing TTA methods, making DART a valuable plug-in tool.

Figures

Figures reproduced from arXiv: 2411.15204 by the authors.

Figure 1
Figure 1. Confusion patterns of BN-adapted classifier due to test-time label distribution shifts. We present class-wise confusion matrices of BN-adapted classifiers, initially trained on class-balanced CIFAR-10 and then tested on CIFAR-10C with two long-tailed distributions (first column). The second column shows confusion patterns on the CIFAR-10 test dataset with only label shifts, while the third to fifth columns display p… view at source ↗
Figure 2
Figure 2. Intermediate time training of DART. At intermediate time, the period between the training and test times, DART trains a prediction refinement module gϕ to correct the inaccurate prediction caused by the class distribution shifts. (left) By sampling the training data from Dirichlet distributions, we generate batches with diverse class distributions. (right) The prediction refinement module gϕ takes the averaged pseud… view at source ↗
Figure 3
Figure 3. We observe performance degrada￾tion of BNAdapt (orange) as the class imbal￾ance ratio ρ increases on long-tailed CIFAR￾10C. DART-applied BNAdapt (green) shows consistently improved performance regard￾less of class imbalance. 2.30 2.32 2.34 2.36 2.38 D(u, 1 |B| i softmax(f (xi))) 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Test accuracy 4.50 4.75 5.00 5.25 5.50 5.75 6.00 6.25 1 |B| i D(u, softmax(f (xi))) 0.2 0.3 0.4 0.5 0.6 0.7… view at source ↗
Figures from the paper (12 more)
Figure 5
Figure 5. Figure 5: Class distribution of PACS and OfficeHome [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: Example of the Dirichlet distribution sampling. IID (i.i.d.) sampling denotes standard uniform sampling. The black dots indicate the class distribution of the sampled batches. The red, blue, and yellow dots represent the class distributions of different class imbalance…
Figure 7
Figure 7. Figure 7: Class distributions of five batches sampled from CIFAR-10 using Dirichlet sampling. These sampled batches [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Illustration of the 4-class Gaussian mixture model from Section D.1. The two lines indicate the Bayes [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 9
Figure 9. Figure 9: Confusion matrices of BN-adapted classifiers on CIFAR-10C-LT with [PITH_FULL_IMAGE:figures/full_fig_p027_9.png]
Figure 10
Figure 10. Figure 10: Confusion matrices of BN-adapted classifiers on CIFAR-10C-LT with [PITH_FULL_IMAGE:figures/full_fig_p028_10.png]
Figure 11
Figure 11. Figure 11: Confusion matrices of BN-adapted classifiers on CIFAR-10C-LT with [PITH_FULL_IMAGE:figures/full_fig_p029_11.png]
Figure 12
Figure 12. Figure 12: Prediction refinement module outputs on CIFAR-10C-LT of [PITH_FULL_IMAGE:figures/full_fig_p030_12.png]
Figure 13
Figure 13. Figure 13: Prediction refinement module outputs on CIFAR-10C-LT of [PITH_FULL_IMAGE:figures/full_fig_p030_13.png]
Figure 14
Figure 14. Figure 14: Prediction refinement module outputs on CIFAR-10C-LT of [PITH_FULL_IMAGE:figures/full_fig_p030_14.png]
Figure 15
Figure 15. Figure 15: Visualizations of ground truth and averaged pseudo label distributions of (left) BNAdapt and (right) DART-applied BNAdapt on CIFAR-10C-LT The averaged refined pseudo label distribution (solid line) generated by the DART-applied BNAdapt closely matches the ground truth…
Figure 16
Figure 16. Figure 16: (a) T-SNE plots of test data with ground truth labels (left) and their predictions (right) for CIFAR-10C-LT [PITH_FULL_IMAGE:figures/full_fig_p035_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 28 canonical work pages

  1. [1]

    Imagenet classification with deep convolutional neural networks

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25, 2012

    2012

  2. [2]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pages 8748–8763. PMLR, 2021

    2021

  3. [3]

    Very deep convolutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014

    arXiv 2014

  4. [4]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018

    arXiv 2018

  5. [5]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in neural information processing systems, 30, 2017

    2017

  6. [6]

    Meta-reinforcement learning robust to distributional shift via model identification and experience relabeling

    Russell Mendonca, Xinyang Geng, Chelsea Finn, and Sergey Levine. Meta-reinforcement learning robust to distributional shift via model identification and experience relabeling. CoRR, abs/2006.07178, 2020

    arXiv 2006

  7. [7]

    Adapting visual category models to new domains

    Kate Saenko, Brian Kulis, Mario Fritz, and Trevor Darrell. Adapting visual category models to new domains. In Proceedings of the 11th European Conference on Computer Vision: Part IV, ECCV’10, page 213–226, Berlin, Heidelberg, 2010. Springer-Verlag

    work page 2010

  8. [8]

    Measuring robustness to natural distribution shifts in image classification

    Rohan Taori, Achal Dave, Vaishaal Shankar, Nicholas Carlini, Benjamin Recht, and Ludwig Schmidt. Measuring robustness to natural distribution shifts in image classification. In H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin, editors, Advances in Neural Information Processing Systems, volume 33, pages 18583–18599. Curran Associates, Inc., 2020

    work page 2020

Show all 56 references
  1. [9]

    Parameter-free online test-time adaptation

    Malik Boudiaf, Romain Mueller, Ismail Ben Ayed, and Luca Bertinetto. Parameter-free online test-time adaptation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8344–8353, 2022

    2022

  2. [10]

    Test time adaptation via conjugate pseudo- labels

    Sachin Goyal, Mingjie Sun, Aditi Raghunathan, and J Zico Kolter. Test time adaptation via conjugate pseudo- labels. In Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho, editors, Advances in Neural Information Processing Systems, 2022

    2022

  3. [11]

    Test-time adaptation via self-training with nearest neighbor information

    Minguk Jang, Sae-Young Chung, and Hye Won Chung. Test-time adaptation via self-training with nearest neighbor information. In The Eleventh International Conference on Learning Representations, 2022

    2022

  4. [12]

    Tent: Fully test-time adaptation by entropy minimization

    Dequan Wang, Evan Shelhamer, Shaoteng Liu, Bruno Olshausen, and Trevor Darrell. Tent: Fully test-time adaptation by entropy minimization. In International Conference on Learning Representations, 2020

    2020

  5. [13]

    Delta: Degradation-free fully test-time adaptation

    Bowen Zhao, Chen Chen, and Shu-Tao Xia. Delta: Degradation-free fully test-time adaptation. In The Eleventh International Conference on Learning Representations, 2022

    2022

  6. [14]

    Eval- uating prediction-time batch normalization for robustness under covariate shift

    Zachary Nado, Shreyas Padhy, D Sculley, Alexander D’Amour, Balaji Lakshminarayanan, and Jasper Snoek. Eval- uating prediction-time batch normalization for robustness under covariate shift. arXiv preprint arXiv:2006.10963, 2020

    2006 arXiv

  7. [15]

    Im- proving robustness against common corruptions by covariate shift adaptation

    Steffen Schneider, Evgenia Rusak, Luisa Eck, Oliver Bringmann, Wieland Brendel, and Matthias Bethge. Im- proving robustness against common corruptions by covariate shift adaptation. Advances in neural information processing systems, 33:11539–11551, 2020

    2020

  8. [16]

    Pseudo-label : The simple and efficient semi-supervised learning method for deep neural networks

    Dong-Hyun Lee. Pseudo-label : The simple and efficient semi-supervised learning method for deep neural networks. In Workshop on challenges in representation learning, ICML, volume 3, 2013

    2013

  9. [17]

    Ods: test-time adaptation in the presence of open-world data shift

    Zhi Zhou, Lan-Zhe Guo, Lin-Han Jia, Dingchu Zhang, and Yu-Feng Li. Ods: test-time adaptation in the presence of open-world data shift. In International Conference on Machine Learning. PMLR, 2023

    2023

  10. [18]

    Note: Robust continual test-time adaptation against temporal correlation

    Taesik Gong, Jongheon Jeong, Taewon Kim, Yewon Kim, Jinwoo Shin, and Sung-Ju Lee. Note: Robust continual test-time adaptation against temporal correlation. Advances in Neural Information Processing Systems, 35:27253–27266, 2022

    2022

  11. [19]

    Towards stable test-time adaptation in dynamic wild world

    Shuaicheng Niu, Jiaxiang Wu, Yifan Zhang, Zhiquan Wen, Yaofo Chen, Peilin Zhao, and Mingkui Tan. Towards stable test-time adaptation in dynamic wild world. In The Eleventh International Conference on Learning Representations, 2023

    2023

  12. [20]

    Learning with noisy labels

    Nagarajan Natarajan, Inderjit S Dhillon, Pradeep K Ravikumar, and Ambuj Tewari. Learning with noisy labels. Advances in neural information processing systems, 26, 2013. 12 Label Distribution Shift-Aware Prediction Refinement for Test-Time Adaptation

    2013

  13. [21]

    Making deep neural networks robust to label noise: A loss correction approach

    Giorgio Patrini, Alessandro Rozza, Aditya Krishna Menon, Richard Nock, and Lizhen Qu. Making deep neural networks robust to label noise: A loss correction approach. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1944–1952, 2017

    1944

  14. [22]

    Clusterability as an alternative to anchor points when learning with noisy labels

    Zhaowei Zhu, Yiwen Song, and Yang Liu. Clusterability as an alternative to anchor points when learning with noisy labels. In Marina Meila and Tong Zhang, editors, Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learnin...

    2021

  15. [23]

    Domain adaptation with invariant representation learning: What transformations to learn? Advances in Neural Information Processing Systems, 34:24791–24803, 2021

    Petar Stojanov, Zijian Li, Mingming Gong, Ruichu Cai, Jaime Carbonell, and Kun Zhang. Domain adaptation with invariant representation learning: What transformations to learn? Advances in Neural Information Processing Systems, 34:24791–24803, 2021

    2021

  16. [24]

    When source-free domain adaptation meets learning with noisy labels

    Li Yi, Gezheng Xu, Pengcheng Xu, Jiaqi Li, Ruizhi Pu, Charles Ling, A Ian McLeod, and Boyu Wang. When source-free domain adaptation meets learning with noisy labels. arXiv preprint arXiv:2301.13381, 2023

    2023 arXiv

  17. [25]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky and Geoffrey Hinton. Learning multiple layers of features from tiny images. Technical Report 0, University of Toronto, Toronto, Ontario, 2009

    2009

  18. [26]

    Benchmarking neural network robustness to common corruptions and perturbations

    Dan Hendrycks and Thomas Dietterich. Benchmarking neural network robustness to common corruptions and perturbations. Proceedings of the International Conference on Learning Representations, 2019

    2019

  19. [27]

    Label shift adapter for test-time adaptation under covariate and label shifts

    Sunghyun Park, Seunghan Yang, Jaegul Choo, and Sungrack Yun. Label shift adapter for test-time adaptation under covariate and label shifts. arXiv preprint arXiv:2308.08810, 2023

    2023 arXiv

  20. [28]

    Improving test-time adaptation via shift-agnostic weight regularization and nearest source prototypes

    Sungha Choi, Seunghan Yang, Seokeon Choi, and Sungrack Yun. Improving test-time adaptation via shift-agnostic weight regularization and nearest source prototypes. In European Conference on Computer Vision, pages 440–458. Springer, 2022

    2022

  21. [29]

    Ttn: A domain-shift aware batch normalization in test-time adaptation

    Hyesu Lim, Byeonggeun Kim, Jaegul Choo, and Sungha Choi. Ttn: A domain-shift aware batch normalization in test-time adaptation. In The Eleventh International Conference on Learning Representations, 2022

    2022

  22. [30]

    Bayesian nonparametric federated learning of neural networks

    Mikhail Yurochkin, Mayank Agarwal, Soumya Ghosh, Kristjan Greenewald, Nghia Hoang, and Yasaman Khazaeni. Bayesian nonparametric federated learning of neural networks. In International conference on machine learning, pages 7252–7261. PMLR, 2019

    2019

  23. [31]

    Do cifar-10 classifiers generalize to cifar-10? arXiv preprint arXiv:1806.00451, 2018

    Benjamin Recht, Rebecca Roelofs, Ludwig Schmidt, and Vaishaal Shankar. Do cifar-10 classifiers generalize to cifar-10? arXiv preprint arXiv:1806.00451, 2018

    2018 arXiv

  24. [32]

    Hospedales

    Da Li, Yongxin Yang, Yi-Zhe Song, and Timothy M. Hospedales. Deeper, broader and artier domain generalization. In IEEE International Conference on Computer Vision, ICCV 2017, Venice, Italy, October 22-29, 2017, pages 5543–5551. IEEE Computer Society, 2017

    2017

  25. [33]

    Deep hashing network for unsupervised domain adaptation

    Hemanth Venkateswara, Jose Eusebio, Shayok Chakraborty, and Sethuraman Panchanathan. Deep hashing network for unsupervised domain adaptation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5018–5027, 2017

    2017

  26. [34]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016

    2016

  27. [35]

    Neural networks: a comprehensive foundation

    Simon Haykin. Neural networks: a comprehensive foundation. Prentice Hall PTR, 1998

    1998

  28. [36]

    Leveraging proxy of training data for test-time adaptation

    Juwon Kang, Nayeong Kim, Kwon Donghyeon, Jungseul Ok, and Suha Kwak. Leveraging proxy of training data for test-time adaptation. In International Conference on Machine Learning (ICML), July 2023

    2023

  29. [37]

    Smote: synthetic minority over-sampling technique

    Nitesh V Chawla, Kevin W Bowyer, Lawrence O Hall, and W Philip Kegelmeyer. Smote: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16:321–357, 2002

    2002

  30. [38]

    Exploratory undersampling for class-imbalance learning

    Xu-Ying Liu, Jianxin Wu, and Zhi-Hua Zhou. Exploratory undersampling for class-imbalance learning. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 39(2):539–550, 2008

    2008

  31. [39]

    Disentangling label distribution for long-tailed visual recognition

    Youngkyu Hong, Seungju Han, Kwanghee Choi, Seokjun Seo, Beomsu Kim, and Buru Chang. Disentangling label distribution for long-tailed visual recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6626–6636, 2021

    2021

  32. [40]

    Long-tail learning via logit adjustment

    Aditya Krishna Menon, Sadeep Jayasumana, Ankit Singh Rawat, Himanshu Jain, Andreas Veit, and Sanjiv Kumar. Long-tail learning via logit adjustment. arXiv preprint arXiv:2007.07314, 2020

    2007 arXiv

  33. [41]

    Rlsbench: Domain adaptation under relaxed label shift

    Saurabh Garg, Nick Erickson, James Sharpnack, Alex Smola, Sivaraman Balakrishnan, and Zachary Chase Lipton. Rlsbench: Domain adaptation under relaxed label shift. In International Conference on Machine Learning, pages 10879–10928. PMLR, 2023. 13 Label Distribution Shift-Aware ...

    2023

  34. [42]

    80 million tiny images: A large data set for nonparametric object and scene recognition

    Antonio Torralba, Rob Fergus, and William T Freeman. 80 million tiny images: A large data set for nonparametric object and scene recognition. IEEE transactions on pattern analysis and machine intelligence, 30(11):1958–1970, 2008

    1958

  35. [43]

    On pitfalls of test-time adaptation

    Hao Zhao, Yuejiang Liu, Alexandre Alahi, and Tao Lin. On pitfalls of test-time adaptation. In International Conference on Machine Learning (ICML), 2023

    2023

  36. [44]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32, 2019

    2019

  37. [45]

    Deep learning using rectified linear units (relu)

    Abien Fred Agarap. Deep learning using rectified linear units (relu). arXiv preprint arXiv:1803.08375, 2018

    2018 arXiv

  38. [46]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  39. [47]

    Distilling the knowledge in a neural network

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015

    2015 arXiv

  40. [48]

    Test-time classifier adjustment module for model-agnostic domain general- ization

    Yusuke Iwasawa and Yutaka Matsuo. Test-time classifier adjustment module for model-agnostic domain general- ization. Advances in Neural Information Processing Systems, 34:2427–2440, 2021

    2021

  41. [49]

    Class relationship embedded learning for source-free unsupervised domain adaptation

    Yixin Zhang, Zilei Wang, and Weinan He. Class relationship embedded learning for source-free unsupervised domain adaptation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7619–7629, 2023

    2023

  42. [50]

    Are anchor points really indispensable in label-noise learning? Advances in neural information processing systems, 32, 2019

    Xiaobo Xia, Tongliang Liu, Nannan Wang, Bo Han, Chen Gong, Gang Niu, and Masashi Sugiyama. Are anchor points really indispensable in label-noise learning? Advances in neural information processing systems, 32, 2019

    2019

  43. [51]

    Dual t: Reducing estimation error for transition matrix in label-noise learning

    Yu Yao, Tongliang Liu, Bo Han, Mingming Gong, Jiankang Deng, Gang Niu, and Masashi Sugiyama. Dual t: Reducing estimation error for transition matrix in label-noise learning. Advances in neural information processing systems, 33:7260–7271, 2020

    2020

  44. [52]

    Dataset distillation by matching training trajectories

    George Cazenavette, Tongzhou Wang, Antonio Torralba, Alexei A Efros, and Jun-Yan Zhu. Dataset distillation by matching training trajectories. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4750–4759, 2022

    2022

  45. [53]

    Universal test-time adaptation through weight ensembling, diversity weighting, and prior correction

    Robert A Marsden, Mario Döbler, and Bin Yang. Universal test-time adaptation through weight ensembling, diversity weighting, and prior correction. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 2555–2565, 2024

    2024

  46. [54]

    An empirical study of training self-supervised vision transformers

    Xinlei Chen, Saining Xie, and Kaiming He. An empirical study of training self-supervised vision transformers. In Proceedings of the IEEE/CVF international conference on computer vision, pages 9640–9649, 2021. 14 Label Distribution Shift-Aware Prediction Refinement for Test-Tim...

    2021

  47. [55]

    mixed domain

    (33) Since d is greater than K generally, the least-square solution for the above equation is W ∗ = {(IK − 1K×1p)µµT (IK − 1K×1p)T }−1(IK − 1K×1p)µµT (IK − 1K×1q)T , (34) with an assumption that (IK − 1K×1p)µµT (IK − 1K×1p)T is invertible. Then, we can observe that W ∗ is dete...

  48. [56]

    This demonstrates the efficiency of DART’s intermediate-time training, significantly enhancing its scalability for practical use

    Notably, training for only 10 epochs achieves comparable results to 50 epochs, with just a 1.8% gap observed for CIFAR-10C-LT withρ = 100. This demonstrates the efficiency of DART’s intermediate-time training, significantly enhancing its scalability for practical use. G.5 DART...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.