Pith. sign in

REVIEW 3 major objections 4 minor 64 references

Efficient Video Dataset Distillation via Cluster-Guided Prototype Blending

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Blending real clips distills video data without costly optimization.

desk verdict Solid construction-based video distillation method with honest ablations, but the headline efficiency claim excludes teacher preparation and candidate inference, so read Fig. 1 as post-teacher construction time only. read the letter →

arxiv 2608.03269 v1 pith:G2GHHU2O submitted 2026-08-04 cs.CV cs.AI

Chongle Ren , Guang Li , Wenbo Huang , Naoki Saito , Takahiro Ogawa , Miki Haseyama This is my paper
classification cs.CVcs.AI
keywords videodatasetdistillationactionrecognitionprototypeblendingteacher-guidedselectioncluster-guidedallocationsoftlabelsconstruction-basedMAE
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

ProtoBlend claims that video dataset distillation does not require iteratively optimizing synthetic videos. Instead, it constructs a compact surrogate set by selecting the most teacher-confident temporal clip from each source video, clustering those clips in teacher feature space, and blending in-cluster prototype–anchor pairs into single stored videos. Those stored videos are trained with mixture-source soft labels derived from the same blending coefficient, coupling input and supervision. On four action-recognition benchmarks, the paper reports accuracy comparable to or better than optimization-based baselines while cutting construction time to under an hour. A sympathetic reader should see this as evidence that useful video surrogates can be built by guided selection and interpolation rather than repeated gradient updates.

What carries the argument

The select–allocate–blend pipeline. Teacher-Guided Temporal Clip Selection retains the clip with the lowest target-class cross-entropy per source video; Cluster-Guided Prototype Allocation runs class-wise K-means on teacher embeddings and assigns one distilled slot per cluster, choosing the lowest-loss clip as prototype and second-lowest as anchor; Blended Sample Construction forms each distilled video as $\tilde{x}_{c,k}=\alpha x^{\star}_{p_{c,k}}+(1-\alpha)x^{\star}_{a_{c,k}}$ and constructs its soft label by mixing the teacher posteriors of the same two clips with the same coefficient $\alpha$. This shared coefficient is the mechanism that keeps the input blend and the supervision consist

What would settle it

Replace the teacher-guided temporal clip selection (Eqs. 3–4) with uniform random clip selection while keeping clustering and blending unchanged, and measure distilled-set accuracy on MiniUCF. If accuracy does not fall to the level of the Random or Herding coreset baselines, the teacher's cross-entropy signal is not the load-bearing component the paper claims; if it does fall, that confirms the teacher's scoring is necessary.

Watch

Extended reading notes

Core claim

The central claim is that a frozen, dataset-specific VideoMAE teacher can carry the entire video distillation pipeline. Its target-class cross-entropy scores select the most discriminative temporal segment from each source video; its embeddings define intra-class modes that allocate the limited videos-per-class budget; and its prediction posteriors, mixed with the same coefficient used for pixel-level blending, supply soft labels aligned with the constructed inputs. Because the distilled set is built from forward inference, class-wise K-means, and direct interpolation, no gradient update is ever applied to the stored videos. The paper reports that this construction-based approach outperforms

Load-bearing premise

The load-bearing premise is that a single frozen teacher's cross-entropy scores, feature embeddings, and prediction posteriors are all reliable enough to guide selection, clustering, and soft-label supervision; if teacher signal degrades, all three stages degrade together and the accuracy advantage disappears.

Editorial extensions

If this is right

  • ProtoBlend's post-teacher construction cost is driven by a single teacher-inference pass; increasing VPC from 1 to 5 barely changes runtime, while optimization-based methods' cost grows with the number of distilled videos.
  • With VPC of 5 or more, cluster-guided allocation adds clear gains over selecting only the most confident clips, indicating that intra-class diversity coverage becomes more valuable than per-sample quality as the budget grows.
  • The accuracy gain transfers across student architectures: the same MiniUCF distilled sets rank first under ConvNet3D, CNN+GRU, and CNN+LSTM, suggesting the constructed videos retain general temporal cues rather than overfitting the evaluation model.
  • Blending and mixture-source soft labels are coupled: with hard labels blending hurts at VPC=5, but soft labels aligned to the blend recover and improve accuracy at both VPC=1 and VPC=5.
  • Random same-class pairing substantially degrades accuracy relative to in-cluster pairing, so the benefit of blending depends on the anchor being from the same teacher-feature cluster as the prototype.

Reading between the lines

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

  • Because construction cost weakly depends on the budget, the cached teacher outputs could be reused to generate multiple compression levels (VPC=1, 5, 10) at near-zero marginal cost, making ProtoBlend practical for rapid hyperparameter search and cross-architecture studies on surrogate sets.
  • The observed plateau on Something-Something V2 and Kinetics-400 suggests that frame-aligned pixel blending is the bottleneck for motion-centric classes; a motion-phase-aware warp or a class-adaptive allocation scheme could recover larger-budget gains where this paper sees diminishing returns.
  • The reliance on a dataset-specific teacher implies ProtoBlend inherits whatever biases the teacher has; testing with a single generalist video model as teacher could make the method dataset-agnostic and cheaper to deploy on new domains.
  • One could probe how sensitive the allocation stage is to the choice of feature space: replacing teacher embeddings with optical-flow or action-tag embeddings for K-means would reveal whether the teacher's appearance-biased representation is the right geometry for defining intra-class modes.
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

3 major / 4 minor

Summary. ProtoBlend proposes a construction-based video dataset distillation method that avoids iterative gradient-based optimization of the stored videos. For each source video, a frozen VideoMAE teacher (fine-tuned on the source dataset) scores R=5 temporal candidates and keeps the one with the lowest target-class cross-entropy (TGS, Eq. 3–4). The selected clips of each class are clustered with K-means in teacher feature space, with one distilled slot per cluster (CGA). Within each cluster, the lowest- and second-lowest-loss clips are chosen as prototype and anchor, then blended via pixel-level interpolation with coefficient α (Eq. 8); the teacher posteriors of the two clips are mixed with the same coefficient to form the soft label (Eq. 9). A student is trained from random initialization on the resulting set using a combined hard/soft loss (Eq. 11–13). The paper reports experiments on MiniUCF, HMDB51, Kinetics-400, and Something-Something V2, comparing with coreset baselines and optimization-based distillation methods. The central claims are that ProtoBlend achieves competitive accuracy and a favorable accuracy–efficiency trade-off without optimizing the distilled videos.

Significance. If the efficiency claim holds, ProtoBlend offers a simple, non-iterative alternative to video dataset distillation, whose optimization cost is a recognized bottleneck. The paper has several strengths: the ablations isolate the contributions of temporal selection, allocation, blending, and soft-label supervision; results are repeated over three student runs with standard deviations; hyperparameter sensitivity is reported; cross-architecture transfer is examined; and the limitations are stated explicitly. The method is falsifiable and easy to reproduce in principle. However, the central efficiency claim is not fully substantiated because the cost of preparing the dataset-specific teacher and running teacher inference over all candidates is excluded from the reported time. Furthermore, on the two larger benchmarks the method is at or below simple coreset baselines, weakening the broad claim of competitive performance. No code or supplementary material is provided, which limits verification.

major comments (3)
  1. [Efficiency Analysis; Fig. 3; Eqs. (3)–(4)] The paper's title and abstract claim "efficient" and a favorable accuracy–efficiency trade-off, but the only time reported is "post-teacher construction time" in Fig. 3. This excludes two large costs: (i) fine-tuning a dataset-specific VideoMAE teacher on the full training set, and (ii) teacher forward inference over R=5 temporal candidates for every source video (Eqs. 3–4). The text even says ProtoBlend shifts cost to "a single teacher-inference stage," yet the teacher is iteratively trained, not just run once. Since the optimization baselines (DM+VDSD, PRISM) do not require such a teacher, the total-cost comparison in Fig. 1 may reverse once teacher preparation is included. Please report wall-clock/GPU-hours for teacher fine-tuning and candidate inference on all datasets, or explicitly re-scope the efficiency claim to the post-teacher construction phase.
  2. [Implementation details; Limitations] ProtoBlend uses a VideoMAE-Base teacher fine-tuned separately on each full source dataset. This teacher supplies clip scoring, feature clustering, and soft labels. The optimization-based baselines (DM, VDSD, PRISM) do not have access to such a strong, dataset-specific teacher. This is not circular, but it makes the accuracy comparison less apples-to-apples: the gains could largely reflect the teacher's pretrained and fine-tuned knowledge rather than the select-allocate-blend mechanism itself. To support the claimed advantage, please ablate the teacher by (a) using a fixed pretrained VideoMAE without fine-tuning, (b) giving the baselines the same teacher-derived soft labels or a similarly strong initialization, or at minimum (c) reporting the teacher's full-dataset accuracy and the cost of obtaining it. The Limitations section acknowledges teacher dependence but does not address the compa
  3. [Table 2; Abstract] The abstract states ProtoBlend demonstrates "competitive accuracy–efficiency trade-off" on four benchmarks, but Table 2 shows that on SSv2 at VPC=5 ProtoBlend (4.3) is below the simple K-center coreset (4.5), and on Kinetics-400 at VPC=5 it is below DM (8.0 vs 9.1). These are not competitive large-scale results, and the paper itself notes limited returns. The abstract and conclusion should qualify the claim to small/medium benchmarks (MiniUCF, HMDB51) or provide additional evidence that the efficiency advantage on these larger datasets offsets the accuracy gap. No efficiency measurements are reported for HMDB51, Kinetics-400, or SSv2, so the trade-off claim is only demonstrated on MiniUCF.
minor comments (4)
  1. [Fig. 1 and Fig. 3] The y-axis "Distillation Time (h)" in Fig. 3 is described in the text as post-teacher time. Please make this explicit in the figure or caption (e.g., "Post-teacher construction time") to avoid misleading readers who compare it directly with the optimization baselines.
  2. [Table 2] The caption or surrounding text should note that the strongest competitor on SSv2 VPC=5 is K-center, a simple coreset method, and that ProtoBlend's advantage over optimization-based methods is within noise on some large-scale settings. This would help calibrate the conclusion.
  3. [Experimental Setup; Table 6] The number of temporal candidates R=5 is fixed and not ablated. Since TGS is the largest contributor in Table 4, a sensitivity analysis on R would clarify how robust the selection stage is to this hyperparameter.
  4. [Implementation details] The paper says "Complete implementation and training hyperparameters are provided in the supplementary material," but no supplementary file is included with the arXiv submission. Please provide code or a detailed hyperparameter table for all datasets and baselines to enable reproduction.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the construction pipeline and the evaluation are decoupled; the central accuracy claim is measured on held-out data with an independent student and an external teacher.

full rationale

I examined the full derivation chain. Clip selection (Eqs. 3–4) uses a frozen, externally published VideoMAE teacher; prototype allocation (Eqs. 5–7) uses K-means on those teacher features; blending (Eq. 8) and soft-label construction (Eq. 9) are matched by design through the shared coefficient α. None of these steps defines the measured outcome (student top-1/top-5 accuracy) in terms of the construction inputs. The student is a randomly initialized ConvNet3D/CNN+GRU/CNN+LSTM trained only on the distilled set, and the teacher is absent at test time, so the evaluation is statistically and architecturally independent of the teacher signal that drives construction. The mixture-source label is a consistency property — the target is aligned to the blended input — not a fitted quantity renamed as a prediction; hyperparameters (α=0.6, λ=0.6, τ=4) are set once and reported with sensitivity curves. The many Li et al. references are related-work background (soft-label distillation, dataset-distillation surveys) and none is load-bearing: no uniqueness theorem is imported from the authors, and the teacher (VideoMAE, Tong et al. 2022) is external. The efficiency analysis measures only post-teacher construction time and the conclusion explicitly narrows the claim to 'lower post-teacher construction cost'; the exclusion of teacher fine-tuning and candidate inference from Fig. 3 is a measurement-scope concern (correctness risk), not circularity of derivation. Per the hard rules, I find no equation-level or citation-level reduction of the claimed result to its inputs, so the appropriate finding is no significant circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The method's contribution is real, but it rests on a small set of hand-set hyperparameters, all decided on MiniUCF. The heavier burden is conceptual: the teacher must be good, and blending must preserve useful information, both of which are assumed rather than derived.

free parameters (4)
  • alpha (blend coefficient) = 0.6
    Selected by sensitivity analysis on MiniUCF at VPC=1 (Fig. 4a), then reused for all datasets and budgets.
  • lambda (soft-label loss weight) = 0.6
    Selected by sensitivity analysis on MiniUCF at VPC=1 (Fig. 4b), then reused everywhere.
  • tau (distillation temperature) = 4
    Fixed in the implementation table; affects soft-label sharpness and gradient scale in Eqs. (9) and (12), not swept.
  • R (temporal candidates per video) = 5
    Choice of candidate count for temporal selection; changes which clips enter the pool and is fixed across experiments.
assumptions (4)
  • domain assumption A fine-tuned VideoMAE teacher's cross-entropy scores and embeddings reliably rank temporal clip quality and capture intra-class similarity.
    Invoked in Eqs. (3)-(7); no mechanism in the method corrects teacher errors, and Limitations says ProtoBlend depends on teacher quality.
  • domain assumption Pixel-level, frame-aligned interpolation of two same-class clips preserves class-discriminative video information.
    Invoked in Eq. (8); Fig. 5 and Limitations acknowledge ghosting and motion-phase mismatch, so the utility of blending is an empirical premise.
  • domain assumption Combining teacher posteriors of the two sources with the same coefficient alpha produces a supervision signal matched to the blended input.
    Invoked in Eq. (9); the paper calls the pair consistently constructed but offers no derivation that the mixture posterior is the true target of the blended video.
  • domain assumption K-means with K equal to the videos-per-class budget distributes the budget across useful intra-class modes.
    Invoked in Cluster-Guided Prototype Allocation; assumes teacher-embedding clusters are the right notion of intra-class variation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Video Dataset Distillation via Cluster-Guided Prototype Blending." pith.science (2026). https://pith.science/paper/G2GHHU2O

@misc{pith2026260803269,
  author       = {Pith},
  title        = {Pith review of: Efficient Video Dataset Distillation via Cluster-Guided Prototype Blending},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G2GHHU2O}},
  note         = {Machine review of arXiv:2608.03269}
}
read the original abstract

Video dataset distillation aims to compress a large video dataset into a compact surrogate set that preserves its training utility. Most existing approaches synthesize condensed videos through iterative optimization, whose cost is amplified by the temporal dimension. Rather than further reducing the number of optimized variables, we investigate whether effective distilled videos can be constructed without gradient-based optimization of the stored videos. Such a construction-based approach must address three challenges: selecting informative temporal segments, covering diverse intra-class variations under a limited videos-per-class budget, and increasing the information carried by each stored sample. To this end, we propose ProtoBlend, an efficient select-allocate-blend framework. First, teacher-guided temporal clip selection retains a high-confidence segment from each source video. Second, cluster-guided prototype allocation partitions the selected clips in the teacher feature space and assigns one distilled slot to each intra-class cluster. Third, each prototype is blended with an in-cluster anchor, while their teacher predictions are combined using the same coefficient to provide mixture-source supervision. Experiments on four trimmed action-recognition benchmarks demonstrate that ProtoBlend achieves a competitive accuracy-efficiency trade-off without iterative optimization of the distilled videos.

Figures

Figures reproduced from arXiv: 2608.03269 by the authors.

Figure 1
Figure 1. Accuracy versus distilled-set construction time on [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of ProtoBlend. For each source video, a frozen fine-tuned VideoMAE teacher selects the temporal candidate [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Accuracy and distillation time on MiniUCF at [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Sensitivity to the blend coefficient α and supervi￾sion weight λ on MiniUCF at VPC=1. inputs, whereas relying entirely on the mixture-source pos￾terior weakens the explicit class-level target. Their combina￾tion retains class identity while adapting the prediction targ…
Figure 6
Figure 6. Figure 6: Qualitative examples on MiniUCF. For each class, three consecutive rows show the prototype, in-cluster anchor, and [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Qualitative examples on HMDB51. For each class, three consecutive rows show the prototype, in-cluster anchor, and [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Qualitative examples on SSv2. For each class, three consecutive rows show the prototype, in-cluster anchor, and [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Qualitative examples on Kinetics-400. For each class, three consecutive rows show the prototype, in-cluster anchor, [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

64 extracted references · 51 canonical work pages

  1. [1]

    Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , pages=

    Dataset Condensation with Distribution Matching , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , pages=

  2. [2]

    Zhang, Hongyi and Cisse, Moustapha and Dauphin, Yann N and Lopez-Paz, David , journal=

    work page

  3. [3]

    Cui, Jiacheng and Bi, Xinyue and Luo, Yaxin and Zhao, Xiaohan and Liu, Jiacheng and Shen, Zhiqiang , booktitle=

    work page

  4. [4]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Dataset Distillation with Neural Characteristic Function: A Minimax Perspective , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    work page

  5. [5]

    Dataset Distillation via the

    Liu, Haoyang and Li, Yijiang and Xing, Tiancheng and Wang, Peiran and Dalal, Vibhu and Li, Luwei and He, Jingrui and Wang, Haohan , booktitle=. Dataset Distillation via the

    work page

  6. [6]

    Cui, Xiao and Qin, Yulei and Zhou, Wengang and Li, Hongsheng and Li, Houqiang , booktitle=

    work page

  7. [7]

    Proceedings of the International Conference on Learning Representations (ICLR) , year=

    Multimodal Dataset Distillation Made Simple by Prototype-Guided Data Synthesis , author=. Proceedings of the International Conference on Learning Representations (ICLR) , year=

    work page

  8. [8]

    and Lawryshyn, Yuri A

    Sajedi, Ahmad and Khaki, Samir and Amjadian, Ehsan and Liu, Lucy Z. and Lawryshyn, Yuri A. and Plataniotis, Konstantinos N. , booktitle=

Show all 64 references
  1. [9]

    IEEE Transactions on Image Processing , year=

    Curriculum Dataset Distillation , author=. IEEE Transactions on Image Processing , year=

  2. [10]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Towards Stable and Storage-Efficient Dataset Distillation: Matching Convexified Trajectory , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  3. [11]

    Transactions on Machine Learning Research , year=

    Dataset Condensation with Color Compensation , author=. Transactions on Machine Learning Research , year=

  4. [12]

    arXiv preprint arXiv:2601.10090 , year=

    Difficulty-guided Sampling: Bridging the Target Gap between Dataset Distillation and Downstream Tasks , author=. arXiv preprint arXiv:2601.10090 , year=

  5. [13]

    Wang, Yi and Li, Kunchang and Li, Xinhao and Yu, Jiashuo and He, Yinan and Chen, Guo and Pei, Baoqi and Zheng, Rongkun and Wang, Zun and Shi, Yansong and others , booktitle=

  6. [14]

    Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) , year=

    Hyperbolic Dataset Distillation , author=. Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) , year=

  7. [15]

    Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) , pages=

    Beyond Neural Scaling Laws: Beating Power Law Scaling via Data Pruning , author=. Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) , pages=

  8. [16]

    Proceedings of the European Conference on Computer Vision (ECCV) , pages=

    Data-to-Model Distillation: A Data-Efficient Learning Framework , author=. Proceedings of the European Conference on Computer Vision (ECCV) , pages=

  9. [17]

    Proceedings of the European Conference on Computer Vision (ECCV) , pages=

    Expanding Language-Image Pretrained Models for General Video Recognition , author=. Proceedings of the European Conference on Computer Vision (ECCV) , pages=

  10. [18]

    Multi-Modal Data-Efficient

    Kong, Lingdong and Xu, Xiang and Ren, Jiawei and Zhang, Wenwei and Pan, Liang and Chen, Kai and Ooi, Wei Tsang and Liu, Ziwei , journal=. Multi-Modal Data-Efficient

  11. [19]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Dataset Distillation by Matching Training Trajectories , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  12. [20]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    On the Diversity and Realism of Distilled Dataset: An Efficient Dataset Distillation Paradigm , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  13. [21]

    Ran, Fengli and Pu, Xiao and Liu, Bo and Bi, Xiuli and Xiao, Bin , booktitle=

  14. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    Emphasizing Discriminative Features for Dataset Distillation in Complex Scenarios , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

  15. [23]

    arXiv preprint arXiv:1811.10959 , year=

    Dataset Distillation , author=. arXiv preprint arXiv:1811.10959 , year=

    arXiv

  16. [24]

    Proceedings of the International Conference on Machine Learning (ICML) , pages=

    Dataset Condensation via Efficient Synthetic-Data Parameterization , author=. Proceedings of the International Conference on Machine Learning (ICML) , pages=

  17. [25]

    Proceedings of the International Conference on Learning Representations (ICLR) , year=

    Dataset Condensation with Gradient Matching , author=. Proceedings of the International Conference on Learning Representations (ICLR) , year=

  18. [26]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

    Arnab, Anurag and Dehghani, Mostafa and Heigold, Georg and Sun, Chen and Lu. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

  19. [27]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Dancing with Still Images: Video Distillation via Static-Dynamic Disentanglement , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  20. [28]

    Choi, Jaehyun and Hur, Jiwan and Han, Gyojin and Yu, Jaemyung and Kim, Junmo , booktitle=

  21. [29]

    International Journal of Computer Vision , volume=

    Selective Volume Mixup for Video Action Recognition , author=. International Journal of Computer Vision , volume=

  22. [30]

    IEEE Transactions on Circuits and Systems for Video Technology , year=

    Video Understanding with Large Language Models: A Survey , author=. IEEE Transactions on Circuits and Systems for Video Technology , year=

  23. [31]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

    Dataset Distillation: A Comprehensive Review , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

  24. [32]

    Artificial Intelligence Review , volume=

    Knowledge Distillation and Dataset Distillation of Large Language Models: Emerging Trends, Challenges, and Future Directions , author=. Artificial Intelligence Review , volume=

  25. [33]

    International Journal of Computer Vision , volume=

    Human Action Recognition and Prediction: A Survey , author=. International Journal of Computer Vision , volume=

  26. [34]

    arXiv preprint arXiv:2012.06567 , year=

    A Comprehensive Study of Deep Video Action Recognition , author=. arXiv preprint arXiv:2012.06567 , year=

    2012 arXiv

  27. [35]

    arXiv preprint arXiv:1708.00489 , year=

    Active Learning for Convolutional Neural Networks: A Core-Set Approach , author=. arXiv preprint arXiv:1708.00489 , year=

    arXiv

  28. [36]

    Yun, Sangdoo and Oh, Seong Joon and Heo, Byeongho and Han, Dongyoon and Kim, Jinhyung , journal=

  29. [37]

    Wang, Kai and Zhao, Bo and Peng, Xiangyu and Zhu, Zheng and Yang, Shuo and Wang, Shuo and Huang, Guan and Bilen, Hakan and Wang, Xinchao and You, Yang , booktitle=

  30. [38]

    Cui, Feng-Qi and Huang, Jinyang and Zhao, Sirui and Guo, Jinglong and Cai, Qifan and Yan, Xin and Liu, Zhi , journal=

  31. [39]

    Proceedings of the International Conference on Learning Representations (ICLR) , year=

    Towards Lossless Dataset Distillation via Difficulty-Aligned Trajectory Matching , author=. Proceedings of the International Conference on Learning Representations (ICLR) , year=

  32. [40]

    Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) , pages=

    Dataset Distillation Using Neural Feature Regression , author=. Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) , pages=

  33. [41]

    Tong, Zhan and Song, Yibing and Wang, Jue and Wang, Limin , booktitle=

  34. [42]

    Soomro, Khurram and Zamir, Amir Roshan and Shah, Mubarak , journal=

  35. [43]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

    Kuehne, Hildegard and Jhuang, Hueihan and Garrote, Est. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

  36. [44]

    Quo Vadis, Action Recognition? A New Model and the

    Carreira, Joao and Zisserman, Andrew , booktitle=. Quo Vadis, Action Recognition? A New Model and the

  37. [45]

    Proceedings of the International Conference on Machine Learning (ICML) , pages=

    Dataset Condensation with Differentiable Siamese Augmentation , author=. Proceedings of the International Conference on Machine Learning (ICML) , pages=

  38. [46]

    Goyal, Raghav and Ebrahimi Kahou, Samira and Michalski, Vincent and Materzynska, Joanna and Westphal, Susanne and Kim, Heuna and Haenel, Valentin and Fruend, Ingo and Yianilos, Peter and Mueller-Freitag, Moritz and others , booktitle=. The

  39. [47]

    arXiv preprint arXiv:2502.05673 , year=

    The Evolution of Dataset Distillation: Toward Scalable and Generalizable Solutions , author=. arXiv preprint arXiv:2502.05673 , year=

    arXiv

  40. [48]

    Proceedings of the International Conference on Machine Learning (ICML) , pages=

    Herding Dynamical Weights to Learn , author=. Proceedings of the International Conference on Machine Learning (ICML) , pages=

  41. [49]

    2009 , publisher=

    Facility Location: Concepts, Models, Algorithms and Case Studies , author=. 2009 , publisher=

    2009

  42. [50]

    Li, Guang and Zhao, Bo and Wang, Tongzhou , title=

  43. [51]

    Proceedings of the IEEE International Conference on Image Processing (ICIP) , pages=

    Soft-Label Anonymous Gastric X-Ray Image Distillation , author=. Proceedings of the IEEE International Conference on Image Processing (ICIP) , pages=

  44. [52]

    Computer Methods and Programs in Biomedicine , volume=

    Compressed Gastric Image Generation Based on Soft-Label Dataset Distillation for Medical Data Sharing , author=. Computer Methods and Programs in Biomedicine , volume=. 2022 , publisher=

    2022

  45. [53]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), Workshop , pages=

    Dataset Distillation for Medical Dataset Sharing , author=. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), Workshop , pages=

  46. [54]

    IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences , year=

    Dataset Distillation using Parameter Pruning , author=. IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences , year=

  47. [55]

    Neural Networks , volume=

    Importance-aware adaptive dataset distillation , author=. Neural Networks , volume=. 2024 , publisher=

    2024

  48. [56]

    Ma, Hongxu and Li, Guang and Wang, Shijie and Zhou, Dongzhan and Sun, Baoli and Ogawa, Takahiro and Haseyama, Miki and Wang, Zhihui , booktitle=

  49. [57]

    arXiv preprint arXiv:2511.17890 , year=

    Decoupled Audio-Visual Dataset Distillation , author=. arXiv preprint arXiv:2511.17890 , year=

  50. [58]

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , pages=

    Generative Dataset Distillation: Balancing Global Structure and Local Details , author=. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , pages=

  51. [59]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops , year=

    Task-Specific Generative Dataset Distillation with Difficulty-Guided Sampling , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops , year=

  52. [60]

    Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) Workshops , year=

    Information-Guided Diffusion Sampling for Dataset Distillation , author=. Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) Workshops , year=

  53. [61]

    Proceedings of the European Conference on Computer Vision (ECCV) Workshops , year=

    Generative Dataset Distillation Based on Diffusion Model , author=. Proceedings of the European Conference on Computer Vision (ECCV) Workshops , year=

  54. [62]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year=

    Dataset Distillation via Vision-Language Category Prototype , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year=

  55. [63]

    Cai, Wenqi and Zou, Yawen and Li, Guang and Gu, Chunzhi and Zhang, Chao , booktitle=

  56. [64]

    arXiv preprint arXiv:2605.07194 , year=

    Closed-Form Linear-Probe Dataset Distillation for Pre-trained Vision Models , author=. arXiv preprint arXiv:2605.07194 , year=

    arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.