pith. sign in

arxiv: 2606.20196 · v2 · pith:2KDNWJZ4new · submitted 2026-06-18 · 💻 cs.CV

Distill Once, Adapt Life-Long: Exploring Dataset Distillation for Continual Test-Time Adaptation

Hyun-Kurl Jang , Jihun Kim , Hyeokjun Kweon , Kuk-Jin Yoon This is my paper

Pith reviewed 2026-07-01 07:06 UTC · model grok-4.3

classification 💻 cs.CV
keywords continual test-time adaptationdataset distillationdomain shiftsynthetic anchorssource-free adaptationlong-term robustnessimage corruption
0
0 comments X

The pith

Distilled synthetic anchors let continual test-time adaptation stay stable over repeated shifts without storing source data.

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

The paper introduces DO-ALL, which performs dataset distillation once before deployment to create a small set of synthetic anchors that summarize the source distribution. During online adaptation, each target sample is matched to its most aligned anchor to supply references for source replay, representation alignment, and regularization inside existing CTTA methods. This addresses the instability that arises in source-free approaches from compounding self-training errors and catastrophic forgetting. The plug-in approach improves results on CIFAR100-C, ImageNet-C, and the CCC benchmark while preserving privacy by never retaining raw source examples.

Core claim

DO-ALL performs dataset distillation prior to deployment to generate a small set of synthetic distilled anchors summarizing the source distribution. During online adaptation, each target sample is matched to its most semantically aligned anchor, enabling stable source replay, representation alignment, and manifold-smoothing regularization within various CTTA frameworks. This enables long-term adaptation without access to the original source dataset, mitigating compounding self-training errors and catastrophic forgetting.

What carries the argument

Matching each target sample to its closest distilled synthetic anchor to supply stable references for replay, alignment, and regularization.

If this is right

  • Existing CTTA algorithms integrate the anchor matching step with no major redesign.
  • Long-term accuracy holds higher across CIFAR100-C, ImageNet-C, and CCC under repeated shifts.
  • Adaptation continues online and label-free while never storing the original source examples.
  • Self-training errors and forgetting are reduced by the external stable signals from the anchors.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same distillation step could support other online adaptation settings where source data cannot be retained for privacy reasons.
  • Distillation objectives could be tuned specifically to preserve long-horizon matching stability rather than single-step accuracy.
  • The anchor-matching idea might transfer to continual learning pipelines that also face storage or licensing limits.

Load-bearing premise

The distilled anchors provide a faithful, stable summary of the source distribution that can be matched to target samples without introducing new compounding errors during long-term adaptation.

What would settle it

A long sequence of domain shifts on one of the paper's benchmarks where accuracy of the DO-ALL version eventually falls to the same level as the plain source-free baseline.

Figures

Figures reproduced from arXiv: 2606.20196 by Hyeokjun Kweon, Hyun-Kurl Jang, Jihun Kim, Kuk-Jin Yoon.

Figure 1
Figure 1. Figure 1: Left: Overview of our DO-ALL framework. Before deployment, we distill the source dataset into a compact set of synthetic anchors. After deployment, each target sample is matched to the anchor that best aligns with its representation. The selected anchors serve as stable reference points that regularize model updates and prevent drift throughout CTTA. Right: Chronological comparison of accuracy on ImageNet-… view at source ↗
Figure 2
Figure 2. Figure 2: Before deployment, we perform DD once on the source dataset to obtain a compact set of source-distilled anchors A˜, where each anchor stores a synthetic input, its soft label under the source model, and its latent feature representation (Sec. 3.1). After deployment, during CTTA, each incoming target sample is matched to its most semantically aligned anchor in feature space, forming the associated anchor se… view at source ↗
Figure 3
Figure 3. Figure 3: Correlation between the quality of the anchor set and its effect on CTTA. The x-axis denotes the validation accuracy obtained by training a classifier solely on the small set, while the y-axis shows the corresponding improvement achieved by integrat￾ing the set into DO-ALL. A positive trend indicates that higher-quality anchors lead to stronger CTTA stabilization, highlighting DO-ALL’s ability to exploit s… view at source ↗
read the original abstract

Continual Test-Time Adaptation (CTTA) aims to maintain model performance under evolving target domains by adapting online without labeled data. However, practical deployments often cannot retain the source dataset due to privacy or licensing constraints, and purely source-free CTTA methods tend to become unstable under long-term distribution shift, suffering from compounding self-training errors and catastrophic forgetting. We introduce DO-ALL (Distill Once, Adapt Life-Long), a plug-and-play framework that revisits source information in a compact and privacy-conscious form via Dataset Distillation (DD). Before deployment, DO-ALL performs DD to produce a small set of synthetic distilled anchors that summarize the source distribution. During adaptation, each target sample is matched with its most semantically aligned anchor, which provides a stable reference for various CTTA via source replay, representation alignment, and manifold-smoothing regularization. DO-ALL can be seamlessly integrated into existing CTTA algorithms, consistently improving long-term robustness across CIFAR100-C, ImageNet-C, and the CCC benchmark. This demonstrates the potential of leveraging DD to enable stable and continuous adaptation without retaining raw source data. The code is available at https://github.com/blue-531/DOALL.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript introduces DO-ALL (Distill Once, Adapt Life-Long), a plug-and-play framework for Continual Test-Time Adaptation (CTTA). It performs dataset distillation once before deployment to produce a small set of synthetic anchors summarizing the source distribution. During adaptation, each target sample is matched to its most semantically aligned anchor to enable source replay, representation alignment, and manifold-smoothing regularization. The approach is presented as integrable into existing CTTA methods without retaining raw source data, with the claim of consistent improvements in long-term robustness on CIFAR100-C, ImageNet-C, and the CCC benchmark. Code is provided at https://github.com/blue-531/DOALL.

Significance. If the empirical results hold, this work is significant because it offers a compact, privacy-preserving mechanism to stabilize long-term CTTA by revisiting source information via distilled anchors, addressing compounding errors and catastrophic forgetting in source-free settings. The public code release is a clear strength supporting reproducibility of the pipeline.

major comments (1)
  1. [Abstract] Abstract: the central claim that DO-ALL 'consistently improving long-term robustness across CIFAR100-C, ImageNet-C, and the CCC benchmark' is asserted without any quantitative results, ablation details, error bars, or baseline comparisons. This is load-bearing for the empirical contribution and must be substantiated with full experimental evidence.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive evaluation of DO-ALL's significance and the code release. We address the single major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that DO-ALL 'consistently improving long-term robustness across CIFAR100-C, ImageNet-C, and the CCC benchmark' is asserted without any quantitative results, ablation details, error bars, or baseline comparisons. This is load-bearing for the empirical contribution and must be substantiated with full experimental evidence.

    Authors: We agree that the abstract, as a concise summary, should include concrete quantitative support for the central claim rather than relying solely on the detailed results in the body. In the revised version we will augment the abstract with key performance numbers (e.g., average accuracy gains on each benchmark relative to the strongest baselines) while retaining the word limit. The full experimental evidence—including tables with error bars, baseline comparisons, and ablations—is already provided in Sections 4 and 5; the revision will simply make the abstract self-contained on this point. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces DO-ALL as a plug-and-play empirical framework that applies dataset distillation once to generate synthetic anchors, then integrates them into existing CTTA methods for replay/alignment/regularization. No equations, derivations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or method description. The central claim is supported by external benchmark results (CIFAR100-C, ImageNet-C, CCC) rather than any internal reduction to inputs by construction. This is a standard engineering contribution with independent empirical validation.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that dataset distillation produces usable anchors and that semantic matching to those anchors supplies stable guidance; these are domain assumptions in machine learning rather than derived results.

free parameters (1)
  • number and selection of distilled anchors
    The size and content of the synthetic set must be chosen to balance compactness against representational fidelity, though specific values are not stated in the abstract.
axioms (1)
  • domain assumption Dataset distillation produces anchors that can be reliably matched to target samples to provide stable references for adaptation.
    Invoked as the basis for the source-replay and alignment steps.

pith-pipeline@v0.9.1-grok · 5756 in / 1178 out tokens · 42193 ms · 2026-07-01T07:06:20.601664+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

81 extracted references · 18 canonical work pages · 5 internal anchors

  1. [1]

    arXiv preprint arXiv:2410.03263 (2024)

    Adachi, K., Yamaguchi, S., Kumagai, A., Hamagami, T.: Test-time adaptation for regression by subspace alignment. arXiv preprint arXiv:2410.03263 (2024)

    work page arXiv 2024

  2. [2]

    In: 2023 IEEE International Conference on Image Processing (ICIP)

    Adachi, K., Yamaguchi, S., Kumagai, A.: Covariance-aware feature alignment with pre-computed source statistics for test-time adaptation to multiple image corrup- tions. In: 2023 IEEE International Conference on Image Processing (ICIP). pp. 800–804. IEEE (2023)

    2023

  3. [3]

    In: Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition

    Boudiaf, M., Mueller, R., Ben Ayed, I., Bertinetto, L.: Parameter-free online test- time adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition. pp. 8344–8353 (2022)

    2022

  4. [4]

    arXiv preprint arXiv:2511.16674 (2025)

    Cazenavette, G., Torralba, A., Sitzmann, V.: Dataset distillation for pre-trained self-supervised vision models. arXiv preprint arXiv:2511.16674 (2025)

    work page arXiv 2025

  5. [5]

    In: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition

    Cazenavette, G., Wang, T., Torralba, A., Efros, A.A., Zhu, J.Y.: Dataset distilla- tion by matching training trajectories. In: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition. pp. 4750–4759 (2022)

    2022

  6. [6]

    Transactions on Machine Learning Research (2023)

    Chakrabarty, G., Sreenivas, M., Biswas, S.: Santa: Source anchoring network and target alignment for continual test time adaptation. Transactions on Machine Learning Research (2023)

    2023

  7. [7]

    In: European Con- ference on Computer Vision

    Choi, S., Yang, S., Choi, S., Yun, S.: Improving test-time adaptation via shift- agnostic weight regularization and nearest source prototypes. In: European Con- ference on Computer Vision. pp. 440–458. Springer (2022)

    2022

  8. [8]

    Advances in Neural Information Processing Systems35, 810–822 (2022)

    Cui, J., Wang, R., Si, S., Hsieh, C.J.: Dc-bench: Dataset condensation benchmark. Advances in Neural Information Processing Systems35, 810–822 (2022)

    2022

  9. [9]

    In: International Conference on Machine Learning

    Cui, J., Wang, R., Si, S., Hsieh, C.J.: Scaling up dataset distillation to imagenet- 1k with constant memory. In: International Conference on Machine Learning. pp. 6565–6590. PMLR (2023)

    2023

  10. [10]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Cui, X., Qin, Y., Zhou, W., Li, H., Li, H.: Optical: Leveraging optimal transport for contribution allocation in dataset distillation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 15245–15254 (2025)

    2025

  11. [11]

    In: 2009 IEEE conference on computer vision and pattern recognition

    Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: Imagenet: A large- scale hierarchical image database. In: 2009 IEEE conference on computer vision and pattern recognition. pp. 248–255. Ieee (2009)

    2009

  12. [12]

    In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Deng, W., Li, W., Ding, T., Wang, L., Zhang, H., Huang, K., Huo, J., Gao, Y.: Ex- ploiting inter-sample and inter-feature relations in dataset distillation. In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 17057–17066 (2024) 16 Jang et al

    2024

  13. [13]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Döbler, M., Marsden, R.A., Yang, B.: Robust mean teacher for continual and gradual test-time adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7704–7714 (2023)

    2023

  14. [14]

    Dong, T., Zhao, B., Lyu, L.: Privacy for free: How does dataset condensation help privacy? In: International Conference on Machine Learning. pp. 5378–5396. PMLR (2022)

    2022

  15. [15]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Du, J., Jiang, Y., Tan, V.Y., Zhou, J.T., Li, H.: Minimizing the accumulated trajectory error to improve dataset distillation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 3749–3758 (2023)

    2023

  16. [16]

    Advances in Neural Information Processing Systems36, 67487–67504 (2023)

    Du, J., Shi, Q., Zhou, J.T.: Sequential subset matching for dataset distillation. Advances in Neural Information Processing Systems36, 67487–67504 (2023)

    2023

  17. [17]

    arXiv preprint arXiv:2310.05773 (2023)

    Guo, Z., Wang, K., Cazenavette, G., Li, H., Zhang, K., You, Y.: Towards loss- less dataset distillation via difficulty-aligned trajectory matching. arXiv preprint arXiv:2310.05773 (2023)

    work page arXiv 2023

  18. [18]

    He,K.,Zhang,X.,Ren,S.,Sun,J.:Deepresiduallearningforimagerecognition.In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016)

    2016

  19. [19]

    Benchmarking Neural Network Robustness to Common Corruptions and Perturbations

    Hendrycks,D.,Dietterich,T.:Benchmarkingneuralnetworkrobustnesstocommon corruptions and perturbations. arXiv preprint arXiv:1903.12261 (2019)

    work page internal anchor Pith review Pith/arXiv arXiv 1903

  20. [20]

    Advances in Neural Infor- mation Processing Systems37, 74211–74232 (2024)

    Jang, H.K., Kim, J., Kweon, H., Yoon, K.J.: Talos: Enhancing semantic scene completion via test-time adaptation on the line of sight. Advances in Neural Infor- mation Processing Systems37, 74211–74232 (2024)

    2024

  21. [21]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Jeong, J., Kwon, H., Kim, M., Yoon, K.J.: Multimodal distribution matching for vision-language dataset distillation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 23072–23082 (2026)

    2026

  22. [22]

    In: Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision

    Jung, S., Lee, J., Kim, N., Shaban, A., Boots, B., Choo, J.: Cafa: Class-aware feature alignment for test-time adaptation. In: Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision. pp. 19060–19071 (2023)

    2023

  23. [23]

    Kang, J., Kim, N., Kwon, D., Ok, J., Kwak, S.: Leveraging proxy of training data for test-time adaptation (2023)

    2023

  24. [24]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Kim, H., Jang, H.K., Yoon, K.J.: Test-time training for lidar semantic segmenta- tion under corruption via geometric inlier discrimination. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 24206– 24216 (2026)

    2026

  25. [25]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Kim, J., Kwon, H., Kweon, H., Jeong, W., Yoon, K.J.: Dc-tta: Divide-and-conquer framework for test-time adaptation of interactive segmentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 23279–23289 (2025)

    2025

  26. [26]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Kim, J., Kwon, H., Kweon, H., Yoon, K.J.: Bootstrapping video semantic segmen- tation model via distillation-assisted test-time adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 10766– 10777 (2026)

    2026

  27. [27]

    Krizhevsky, A., Hinton, G., et al.: Learning multiple layers of features from tiny images (2009)

    2009

  28. [28]

    arXiv preprint arXiv:2402.08712 (2024)

    Lee,D.,Yoon,J.,Hwang,S.J.:Becotta:Input-dependentonlineblendingofexperts for continual test-time adaptation. arXiv preprint arXiv:2402.08712 (2024)

    work page arXiv 2024

  29. [29]

    In: International Conference on Machine Learning

    Lee, S., Chun, S., Jung, S., Yun, S., Yoon, S.: Dataset condensation with con- trastive signals. In: International Conference on Machine Learning. pp. 12352– 12364. PMLR (2022) DO-ALL 17

    2022

  30. [30]

    arXiv preprint arXiv:2406.18561 (2024)

    Lee, Y., Chung, H.W.: Selmatch: Effectively scaling up dataset distillation via selection-based initialization and partial updates by trajectory matching. arXiv preprint arXiv:2406.18561 (2024)

    work page arXiv 2024

  31. [31]

    IEEE Transactions on Pattern Analysis and Machine Intelligence46(1), 17–32 (2023)

    Lei, S., Tao, D.: A comprehensive survey of dataset distillation. IEEE Transactions on Pattern Analysis and Machine Intelligence46(1), 17–32 (2023)

    2023

  32. [32]

    In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision

    Li, H., Zhou, Y., Gu, X., Li, B., Wang, W.: Diversity-enhanced distribution align- ment for dataset distillation. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision. pp. 3747–3756 (2025)

    2025

  33. [33]

    arXiv preprint arXiv:2505.24623 (2025)

    Li, W., Li, G., Maeda, K., Ogawa, T., Haseyama, M.: Hyperbolic dataset distilla- tion. arXiv preprint arXiv:2505.24623 (2025)

    work page arXiv 2025

  34. [34]

    Revisiting Batch Normalization For Practical Domain Adaptation

    Li, Y., Wang, N., Shi, J., Liu, J., Hou, X.: Revisiting batch normalization for practical domain adaptation. arXiv preprint arXiv:1603.04779 (2016)

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  35. [35]

    arXiv preprint arXiv:2505.13300 (2025)

    Li, Z., Zhong, X., Khaki, S., Liang, Z., Zhou, Y., Shi, M., Wang, Z., Zhao, X., Zhao, W., Qin, Z., et al.: Dd-ranking: Rethinking the evaluation of dataset distillation. arXiv preprint arXiv:2505.13300 (2025)

    work page arXiv 2025

  36. [36]

    In: International conference on machine learning

    Liang, J., Hu, D., Feng, J.: Do we really need to access the source data? source hy- pothesis transfer for unsupervised domain adaptation. In: International conference on machine learning. pp. 6028–6039. PMLR (2020)

    2020

  37. [37]

    arXiv preprint arXiv:2302.05155 (2023)

    Lim, H., Kim, B., Choo, J., Choi, S.: Ttn: A domain-shift aware batch normaliza- tion in test-time adaptation. arXiv preprint arXiv:2302.05155 (2023)

    work page arXiv 2023

  38. [38]

    In: ICLR (2026)

    Lim, T., Hwang, J.W., Lee, K.: When and where to reset matters for long-term test-time adaptation. In: ICLR (2026)

    2026

  39. [39]

    In: European Conference on Computer Vision

    Liu, D., Gu, J., Cao, H., Trinitis, C., Schulz, M.: Dataset distillation by automatic training trajectories. In: European Conference on Computer Vision. pp. 334–351. Springer (2024)

    2024

  40. [40]

    arXiv preprint arXiv:2311.18531 (2023)

    Liu, H., Li, Y., Xing, T., Dalal, V., Li, L., He, J., Wang, H.: Dataset distillation via the wasserstein metric. arXiv preprint arXiv:2311.18531 (2023)

    work page arXiv 2023

  41. [41]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Liu, J., Xu, R., Yang, S., Zhang, R., Zhang, Q., Chen, Z., Guo, Y., Zhang, S.: Continual-mae: Adaptive distribution masked autoencoders for continual test-time adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 28653–28663 (2024)

    2024

  42. [42]

    IEEE Transactions on Image Processing (2025)

    Ma, Z., Cao, A., Yang, F., Gong, Y., Wei, X.: Curriculum dataset distillation. IEEE Transactions on Image Processing (2025)

    2025

  43. [43]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Marsden, R.A., Döbler, M., Yang, B.: 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. pp. 2555– 2565 (2024)

    2024

  44. [44]

    arXiv preprint arXiv:2011.00050 (2020)

    Nguyen, T., Chen, Z., Lee, J.: Dataset meta-learning from kernel ridge-regression. arXiv preprint arXiv:2011.00050 (2020)

    work page arXiv 2011

  45. [45]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Ni, C., Lyu, F., Tan, J., Hu, F., Yao, R., Zhou, T.: Maintaining consistent inter- class topology in continual test-time adaptation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 15319–15328 (2025)

    2025

  46. [46]

    In: International conference on machine learning

    Niu, S., Wu, J., Zhang, Y., Chen, Y., Zheng, S., Zhao, P., Tan, M.: Efficient test- time model adaptation without forgetting. In: International conference on machine learning. pp. 16888–16905. PMLR (2022)

    2022

  47. [47]

    arXiv preprint arXiv:2302.12400 , year =

    Niu, S., Wu, J., Zhang, Y., Wen, Z., Chen, Y., Zhao, P., Tan, M.: Towards sta- ble test-time adaptation in dynamic wild world. arXiv preprint arXiv:2302.12400 (2023)

    work page arXiv 2023

  48. [48]

    Advances in Neural Information Processing Systems36, 39915–39935 (2023) 18 Jang et al

    Press, O., Schneider, S., Kümmerer, M., Bethge, M.: Rdumb: A simple approach that questions our progress in continual test-time adaptation. Advances in Neural Information Processing Systems36, 39915–39935 (2023) 18 Jang et al

    2023

  49. [49]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Sajedi, A., Khaki, S., Amjadian, E., Liu, L.Z., Lawryshyn, Y.A., Plataniotis, K.N.: Datadam: Efficient dataset distillation with attention matching. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 17097–17107 (2023)

    2023

  50. [50]

    Advances in neural information processing systems37, 99161–99201 (2024)

    Shao, S., Zhou, Z., Chen, H., Shen, Z.: Elucidating the design space of dataset condensation. Advances in neural information processing systems37, 99161–99201 (2024)

    2024

  51. [51]

    In: Proceedings of the Computer Vision and Pat- tern Recognition Conference

    Shen, Z., Sherif, A., Yin, Z., Shao, S.: Delt: A simple diversity-driven earlylate training for dataset distillation. In: Proceedings of the Computer Vision and Pat- tern Recognition Conference. pp. 4797–4806 (2025)

    2025

  52. [52]

    In: International Conference on Artificial In- telligence and Statistics

    Shin, S., Bae, H., Shin, D., Joo, W., Moon, I.C.: Loss-curvature matching for dataset selection and condensation. In: International Conference on Artificial In- telligence and Statistics. pp. 8606–8628. PMLR (2023)

    2023

  53. [53]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Song, J., Lee, J., Kweon, I.S., Choi, S.: Ecotta: Memory-efficient continual test- time adaptation via self-distilled regularization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 11920–11929 (2023)

    2023

  54. [54]

    In: 2023 International Joint Conference on Neural Networks (IJCNN)

    Song, R., Liu, D., Chen, D.Z., Festag, A., Trinitis, C., Schulz, M., Knoll, A.: Feder- ated learning via decentralized dataset distillation in resource-constrained edge en- vironments. In: 2023 International Joint Conference on Neural Networks (IJCNN). pp. 1–10. IEEE (2023)

    2023

  55. [55]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Su, Z., Guo, J., Yao, K., Yang, X., Wang, Q., Huang, K.: Unraveling batch normal- ization for realistic test-time adaptation. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 38, pp. 15136–15144 (2024)

    2024

  56. [56]

    In: ICLR 2024 Workshop on Data-centric Machine Learning Research (DMLR) (2024)

    Sun, P., Shi, B., Shang, X., Lin, T.: Information compensation: A fix for any-scale dataset distillation. In: ICLR 2024 Workshop on Data-centric Machine Learning Research (DMLR) (2024)

    2024

  57. [57]

    Tent: Fully Test-time Adaptation by Entropy Minimization

    Wang, D., Shelhamer, E., Liu, S., Olshausen, B., Darrell, T.: Tent: Fully test-time adaptation by entropy minimization. arXiv preprint arXiv:2006.10726 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2006

  58. [58]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, K., Zhao, B., Peng, X., Zhu, Z., Yang, S., Wang, S., Huang, G., Bilen, H., Wang, X., You, Y.: Cafe: Learning to condense dataset by aligning features. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 12196–12205 (2022)

    2022

  59. [59]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, Q., Fink, O., Van Gool, L., Dai, D.: Continual test-time domain adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7201–7211 (2022)

    2022

  60. [60]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Wang, S., Yang, Y., Liu, Z., Sun, C., Hu, X., He, C., Zhang, L.: Dataset distillation with neural characteristic function: A minmax perspective. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 25570–25580 (2025)

    2025

  61. [61]

    Dataset Distillation

    Wang, T., Zhu, J.Y., Torralba, A., Efros, A.A.: Dataset distillation. arXiv preprint arXiv:1811.10959 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  62. [62]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Wang, Y., Hong, J., Cheraghian, A., Rahman, S., Ahmedt-Aristizabal, D., Peters- son, L., Harandi, M.: Continual test-time domain adaptation via dynamic sample selection. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 1701–1710 (2024)

    2024

  63. [63]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Xie,S.,Girshick,R.,Dollár,P.,Tu,Z.,He,K.:Aggregatedresidualtransformations for deep neural networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1492–1500 (2017)

    2017

  64. [64]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Xiong, Y., Wang, R., Cheng, M., Yu, F., Hsieh, C.J.: Feddm: Iterative distribution matching for communication-efficient federated learning. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 16323– 16332 (2023) DO-ALL 19

    2023

  65. [65]

    Advances in Neural Information Processing Systems36, 67625–67642 (2023)

    Yang, E., Shen, L., Wang, Z., Liu, T., Guo, G.: An efficient dataset condensation plugin and its application to continual learning. Advances in Neural Information Processing Systems36, 67625–67642 (2023)

    2023

  66. [66]

    In: European Conference on Computer Vision

    Yang, S., Cheng, S., Hong, M., Fan, H., Wei, X., Liu, S.: Neural spectral decompo- sition for dataset distillation. In: European Conference on Computer Vision. pp. 275–290. Springer (2024)

    2024

  67. [67]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Yang, X., Chen, X., Li, M., Wei, K., Deng, C.: A versatile framework for contin- ual test-time domain adaptation: Balancing discriminability and generalizability. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 23731–23740 (2024)

    2024

  68. [68]

    Advances in Neural Information Processing Systems36, 73582–73603 (2023)

    Yin, Z., Xing, E., Shen, Z.: Squeeze, recover and relabel: Dataset condensation at imagenet scale from a new perspective. Advances in Neural Information Processing Systems36, 73582–73603 (2023)

    2023

  69. [69]

    In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition

    Yuan, L., Xie, B., Li, S.: Robust test-time adaptation in dynamic scenarios. In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition. pp. 15922–15932 (2023)

    2023

  70. [70]

    Zhang, H., Li, S., Lin, F., Wang, W., Qian, Z., and Ge, S

    Zhang,H.,Li,S.,Lin,F.,Wang,W.,Qian,Z.,Ge,S.:Dance:Dual-viewdistribution alignment for dataset condensation. arXiv preprint arXiv:2406.01063 (2024)

    work page arXiv 2024

  71. [71]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Zhang, H., Li, S., Wang, P., Zeng, D., Ge, S.: M3d: Dataset condensation by minimizing maximum mean discrepancy. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 38, pp. 9314–9322 (2024)

    2024

  72. [72]

    mixup: Beyond Empirical Risk Minimization

    Zhang, H., Cisse, M., Dauphin, Y.N., Lopez-Paz, D.: mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  73. [73]

    Advances in neural information processing systems35, 38629–38642 (2022)

    Zhang, M., Levine, S., Finn, C.: Memo: Test time robustness via adaptation and augmentation. Advances in neural information processing systems35, 38629–38642 (2022)

    2022

  74. [74]

    In: Proceedings of the International Conference on Learning Representations (ICLR) (2025)

    Zhang, X., Du, J., Liu, P., Zhou, J.T.: Breaking class barriers: Efficient dataset distillation via inter-class feature compensator. In: Proceedings of the International Conference on Learning Representations (ICLR) (2025)

    2025

  75. [75]

    In: International Conference on Machine Learning

    Zhao, B., Bilen, H.: Dataset condensation with differentiable siamese augmenta- tion. In: International Conference on Machine Learning. pp. 12674–12685. PMLR (2021)

    2021

  76. [76]

    In: Proceed- ings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Zhao, B., Bilen, H.: Dataset condensation with distribution matching. In: Proceed- ings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 6514–6523 (2023)

    2023

  77. [77]

    Dataset condensation with gradientmatching

    Zhao, B., Mopuri, K.R., Bilen, H.: Dataset condensation with gradient matching. arXiv preprint arXiv:2006.05929 (2020)

    work page arXiv 2006

  78. [78]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Zhao, G., Li, G., Qin, Y., Yu, Y.: Improved distribution matching for dataset condensation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7856–7865 (2023)

    2023

  79. [79]

    In: Proceed- ings of the Computer Vision and Pattern Recognition Conference

    Zhong, W., Tang, H., Zheng, Q., Xu, M., Hu, Y., Guan, W.: Towards stable and storage-efficient dataset distillation: Matching convexified trajectory. In: Proceed- ings of the Computer Vision and Pattern Recognition Conference. pp. 25581–25589 (2025)

    2025

  80. [80]

    Advances in Neural Information Processing Systems35, 9813–9827 (2022)

    Zhou, Y., Nezhadarya, E., Ba, J.: Dataset distillation using neural feature regres- sion. Advances in Neural Information Processing Systems35, 9813–9827 (2022)

    2022

Showing first 80 references.