Pith. sign in

REVIEW 3 major objections 4 minor 28 references

CLEAR: Class-wise Expert Aggregation with Structured Sampling for Long-Tailed Classification

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Modeling expert reliability per class improves few-shot accuracy while keeping overall accuracy competitive on long-tailed benchmarks.

desk verdict A genuinely new class-wise reliability weighting for long-tailed ensembles, with strong few-shot gains, but the experiments never isolate the proposed aggregation rule from the other moving parts. read the letter →

arxiv 2608.11287 v1 pith:AVGW3NEX submitted 2026-08-11 cs.CV cs.AIcs.LG

Gawon Lim This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords long-tailedrecognitionclass-wisetrustexpertaggregationstructuredsamplingproductofexpertsfew-shotaccuracyensemblelearningreliabilityestimation
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

Long-tailed classifiers are reliably accurate on frequent classes but unreliable on rare ones, and existing multi-expert methods typically assign each expert a global weight or route samples rather than asking which expert to trust for each class. This paper proposes CLEAR, a modular ensemble that trains several experts on progressively clipped versions of the training set and then gives each expert a separate trust score per class, estimated as a smoothed class-wise precision. Predictions are combined with a class-wise product-of-experts rule, so a class can lean on the expert that handles it best. On CIFAR-100-LT, ImageNet-LT, and Places-LT, CLEAR keeps overall accuracy competitive while improving few-shot accuracy, with the largest tail gains on CIFAR-100-LT (41.25% few-shot) and Places-LT (40.82%). The paper's point is that class-wise expert reliability is a useful design principle for long-tailed ensembles.

What carries the argument

The load-bearing machinery is the class-wise trust score q_{m,c}, defined as the posterior mean of a Beta-binomial model over expert m's precision when predicting class c: q_{m,c} = (alpha_0 + n_{m,c}) / (alpha_0 + beta_0 + N_{m,c}), where N_{m,c} is the number of times expert m predicts class c and n_{m,c} is the number of those that are correct. These scores are normalized across experts by a softmax with sharpness tau, yielding weights w_{m,c}, and the final logit for class c is S_c(x) = sum_m w_{m,c} log p_m(c|x). This class-wise generalized product-of-experts rule lets different experts dominate different classes, and tau interpolates between geometric-mean aggregation and hard selection of the most trusted expert per class.

What would settle it

Train CLEAR on CIFAR-100-LT but estimate every class-wise trust score from held-out predictions (e.g., by splitting each class's samples before training, or using out-of-bag predictions), then re-measure few-shot accuracy. If the few-shot gains over single experts and over SADE shrink to near zero, the in-bag optimism is the source of the reported tail improvement.

Watch

Extended reading notes

Core claim

CLEAR's central claim is that expert reliability in long-tailed classification is inherently class-dependent, so ensemble weighting should be class-wise rather than global. To instantiate this, the paper introduces structured sampling: a schedule of clipping thresholds (exponential decay by default) generates sub-training sets that move from the original long-tailed distribution toward balanced ones, with every class preserved in every subset, and one expert is trained per subset. For each expert and class, a trust score is the posterior mean of a Beta-binomial model of class-wise precision, and these scores are sharpened with a softmax temperature and used as exponents in a generalized product-of-experts aggregation. The authors report that this scheme yields competitive overall accuracy and particularly strong few-shot performance on three benchmarks, supporting the claim that class-wise reliability-aware aggregation is a useful principle for long-tailed ensemble learning.

Load-bearing premise

For classes whose samples are all used to train an expert, the trust score is computed from the expert's in-bag training predictions, which can be optimistically biased for rare classes and may inflate the reported few-shot advantage.

Editorial extensions

If this is right

  • Adding class-wise trust weighting to a standard long-tailed objective, such as Balanced Softmax, improves few-shot accuracy without retraining or changing the loss.
  • The sharpness parameter tau lets practitioners slide between uniform averaging and per-class hard expert selection, so the aggregation can be tuned to how noisy the trust estimates are.
  • Strong few-shot gains appear on both object-centric (CIFAR-100-LT) and scene-centric (Places-LT) benchmarks, suggesting the principle transfers across dataset types.
  • Most of the accuracy gain comes from the first 5–8 experts, so the ensemble can be kept compact for deployment.

Reading between the lines

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

  • Beyond the paper, the same class-wise trust mechanism could be applied to any collection of pretrained models, not only the clipped-distribution experts, as a post-hoc ensemble weighting that requires no retraining.
  • The in-bag trust estimation for rare classes is the most likely source of optimistic few-shot gains; replacing it with out-of-bag or cross-validated estimates would test whether the reported tail improvements are real or partly an artifact.
  • The Beta-binomial trust model could be extended to estimate uncertainty in the trust scores themselves, allowing the aggregation to down-weight classes where trust is uncertain as well as classes where trust is low.
  • One could test the same design principle in other imbalanced domains, such as medical imaging or fraud detection, where rare classes are consequential and multiple models are already available.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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. This paper proposes CLEAR, a multi-expert ensemble for long-tailed classification. Using exponential decay clipping, the method trains M experts on progressively more balanced subsets of the original training set. For each expert and class it estimates a trust score q_{m,c} as the posterior mean of a Beta-binomial model of class-wise precision, based on held-out predictions when the class has more than the clipping threshold and on in-bag predictions otherwise. Trust scores are normalized across experts with a softmax temperature τ, and test predictions are aggregated in log space as S_c(x)=Σ_m w_{m,c} log p_m(c|x). The method is evaluated on CIFAR-100-LT, ImageNet-LT, and Places-LT with several backbones, reporting competitive overall accuracy and especially strong few-shot accuracy on CIFAR-100-LT (41.25% few-shot) and Places-LT (40.82%).

Significance. The central idea—that expert reliability in long-tailed recognition should be modeled per class rather than globally—is timely and plausible, and the framework is modular and broadly evaluated. The paper is honest about its limitations, including the in-bag trust bias and computational overhead. If the reported gains are attributable to the class-wise trust weighting, this would be a useful design principle for ensemble-based long-tailed classification. However, the current experimental design does not isolate that mechanism, and the theoretical derivation of the aggregation rule is not fully sound. The contribution is therefore conditional on additional controlled experiments.

major comments (3)
  1. [5.3 (Figure 4), 4.2 (Table 2)] No experiment isolates the class-wise trust weighting from the rest of the pipeline. Figure 4 varies τ within the proposed class-wise scheme; τ=0 corresponds to uniform per-expert geometric-mean aggregation, not to a global-trust or simple-average ensemble built from the same experts. Consequently, the few-shot gains in Table 2 (e.g., 37.45 for CLEAR(BSM) vs. 33.9 for SADE) could in principle be produced by the structured expert set or by the external BCL/LA components rather than by the class-wise reliability estimates. I request an ablation that fixes the expert set and compares (i) uniform aggregation, (ii) global expert weights (e.g., softmax of overall held-out precision), and (iii) the proposed class-wise weights, both with and without BCL/LA.
  2. [3.3 and Algorithm 1, Section 6] The trust scores for the classes that drive the claimed few-shot advantage are estimated from in-bag predictions. As Algorithm 1 (lines 6-8) states, for classes with n_c ≤ T_m no held-out samples remain, so q_{m,c} is computed from training-side predictions. Section 6 explicitly acknowledges that this 'may lead to optimistic trust estimates, particularly for rare classes.' Because the headline few-shot accuracy is measured on exactly those rare classes, this bias is load-bearing: an expert that memorized tail training examples can receive inflated trust and dominate the aggregation at test time. To support the central claim, the authors should quantify this bias, for example by comparing in-bag and held-out trust estimates on classes with n_c > T_m, or by adding a cross-validation-based trust estimation variant.
  3. [3.2, Eqs. (7)-(9)] The gPoE derivation is not internally consistent. Eq. (1) defines θ_{m,c} as a probability of correctness, and Eq. (7) uses it as an exponent on p_m(c|x). In Eq. (9) this exponent is replaced by w_{m,c}, a softmax-normalized weight that sums to one across experts, which is a different quantity. The text says 'replacing θ_{m,c} with the normalized trust w_{m,c}' but offers no justification for why a normalized score should enter the exponent of a product-of-experts posterior. As written, Eq. (10) is a heuristic class-wise log-linear ensemble; that is acceptable as a method, but the Bayesian/gPoE motivation should be either corrected or explicitly downgraded to a heuristic.
minor comments (4)
  1. [4.1] 'Macro Accuracy' is used throughout Section 5 and Figures 2-6 but is not defined in the evaluation metrics; define it explicitly, presumably as mean per-class accuracy.
  2. [Table 6] 'Uniform Interval', 'Quantile Selection', and 'K-means Clustering' are used without a definition in Section 3.3 or the experimental setup; add a sentence describing each schedule.
  3. [Tables 2 and 3] Tables 2 and 3 report different numbers for CLEAR(BSM) on CIFAR-100-LT with ResNet-32 (53.09 vs. 52.70 overall, 37.45 vs. 36.53 few-shot) because Table 2 uses δ=0.95 and Table 3 uses δ=0.9; state this explicitly to prevent reader confusion.
  4. [3.1, Eq. (5)] The 'without smoothing' limit should be written as α0,β0→0, since α0=β0=0 makes the posterior mean undefined for N_{m,c}=0.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: CLEAR's trust scores are estimated from training-side statistics and never from test labels, so the claimed few-shot gains are empirical outcomes rather than construction artifacts.

full rationale

The paper's derivation chain is self-contained in the sense that no equation reduces to its own output. The class-wise trust score q_{m,c} in Eq. (5) is defined as a Beta-posterior mean of class-wise precision, estimated from counts n_{m,c} and N_{m,c} obtained either from held-out samples excluded by clipping (Algorithm 1, line 6) or, for rare classes, from training-side in-bag predictions (line 7). These estimates depend only on training data and expert predictions, not on test labels. The aggregation rule in Eqs. (9)-(10) weights expert log-probabilities by softmax-normalized trust scores, which is a defined heuristic rather than a derivation forced by the data. The reported few-shot accuracy is measured on a held-out balanced test set, so it is not a fitted quantity. The acknowledged limitation in Section 6 that in-bag estimates 'may lead to optimistic trust estimates, particularly for rare classes' is an inferential bias concern, not circularity: an optimistic weight could harm or help test accuracy, but nothing in the construction guarantees the claimed result. There are no self-citations, no imported uniqueness theorems, and no fitted-parameter-renamed-as-prediction steps. The skeptic concern that no ablation isolates the class-wise weighting is a legitimate experimental attribution issue, but it does not constitute circularity under the definition used here. Therefore the appropriate score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or structural entities. Its core axiomatic load consists of modeling assumptions: class-wise precision as reliability, a Beta-Binomial model for correctness counts, the conditional-independence approximation in gPoE, and the validity of in-bag trust estimates for rare classes. The main free parameters are the ensemble sharpness tau, the clipping decay delta, the Beta prior constants, the number of experts, and two optional module weights (LA, BCL) tuned per dataset.

free parameters (6)
  • tau (sharpness) = 2.0 (CIFAR, ImageNet), 1.0 (Places-LT)
    Controls how strongly class-wise weights concentrate on the most reliable expert (Eq. 6); chosen per dataset, affects accuracy.
  • delta (decay rate) = 0.95 (CIFAR Table 2), 0.9 (CIFAR Table 3 and ImageNet-LT), 0.6 (Places-LT)
    Controls how quickly clipping thresholds decrease across stages (Eq. 15); tuned per dataset and directly changes expert diversity and quality.
  • alpha0, beta0 (Beta prior) = 1, 1
    Smoothing parameters in the Beta posterior (Eq. 4); fixed by hand, not fitted. They affect trust estimates for small N.
  • M (number of stages/experts) = 15 for all datasets
    Number of experts; ablation shows gains saturate by 5-8 stages, yet 15 is used in main tables.
  • alpha_LA (logit adjustment weight) = 0.1 (CIFAR, Places), 0.15 (ImageNet-LT)
    Optional inference-time post-hoc logit adjustment (Eq. 18) from the LA baseline; tuned per dataset.
  • lambda_BCL (Balanced Contrastive Learning weight) = 0.1 (CIFAR, ImageNet), 1.2 (Places)
    Weight of auxiliary contrastive loss (Eq. 17) inherited from BCL baseline; tuned per dataset.
assumptions (4)
  • domain assumption Class-wise precision is a valid proxy for test-time reliability of an expert on a class.
    Eq. (1) defines theta_{m,c} as probability correct given prediction c; the paper relies on this being estimable from training-side statistics and stable under Beta smoothing.
  • domain assumption The product-of-experts conditional independence approximation (Eq. 8) is adequate for classification.
    The paper itself states this is an approximation; the aggregation rule in Eq. (9) is motivated by it rather than derived from it.
  • ad hoc to paper The Beta-Binomial model (Eqs. 2-3) correctly describes prediction correctness counts.
    Chosen for closed-form posterior; independence of prediction outcomes across samples is assumed without testing.
  • domain assumption Trust scores computed from in-bag predictions for rare classes remain informative for test-time weighting.
    Algorithm 1 uses in-bag predictions when n_c <= T_m; the authors acknowledge optimism but assume it does not invert the expert ranking.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CLEAR: Class-wise Expert Aggregation with Structured Sampling for Long-Tailed Classification." pith.science (2026). https://pith.science/paper/AVGW3NEX

@misc{pith2026260811287,
  author       = {Pith},
  title        = {Pith review of: CLEAR: Class-wise Expert Aggregation with Structured Sampling for Long-Tailed Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AVGW3NEX}},
  note         = {Machine review of arXiv:2608.11287}
}
read the original abstract

Long-tailed classification poses a reliability challenge because models trained on imbalanced data are unevenly reliable across frequent and underrepresented classes. While existing methods address imbalance through re-balancing, adjustment, representation learning, or multi-expert modeling, they rarely estimate which expert should be trusted for each class. This paper proposes CLEAR (Class-wise reLiability-aware Expert Aggregation for long-tailed Recognition), a modular ensemble framework for long-tailed classification. CLEAR generates diverse experts through threshold-based structured sampling while preserving the full label space, then estimates a class-wise trust score for each expert using a smoothed class-wise precision formulation. During inference, expert predictions are combined through class-wise generalized product-of-experts aggregation, allowing different experts to be emphasized for different classes. Experiments on CIFAR-100-LT, ImageNet-LT, and Places-LT across multiple backbones show that CLEAR achieves competitive overall accuracy and particularly strong few-shot performance. These results support class-wise expert reliability as a useful design principle for long-tailed ensemble learning.

Figures

Figures reproduced from arXiv: 2608.11287 by the authors.

Figure 1
Figure 1. Overview of CLEAR. (Left) Threshold-based structured sampling generates multiple sub-training sets with varying imbalance levels by progressively clipping class frequencies. This process produces diverse expert models specialized in different regions of the label distribution. (Right) The predictions of these experts are aggregated using class-wise trust weights derived from precision. The final prediction is obtain… view at source ↗
Figure 2
Figure 2. Macro Accuracy and Few-shot Accuracy vs. ensemble stage m. All results use BSM loss with τ = 2.0. Shaded regions denote ±1 standard deviation across seeds. (BSM/BCL/LA) variant achieves 41.25% Few-shot accuracy, showing a clear advantage over strong recent baselines. On ImageNet-LT, CLEAR (BSM/BCL/LA) reaches 58.43% overall accuracy, while the BSM/LA variant achieves 43.48% Few-shot accuracy, indicating that inferen… view at source ↗
Figure 3
Figure 3. Macro Accuracy and Few-shot Accuracy vs. decay rate δ on CIFAR-100-LT. All results use BSM loss. 5.2 Effect of the Clipping Strategy [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Macro Accuracy and Few-shot Accuracy as a function of τ . All results use BSM loss at the final ensemble stage. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 BCL weight ( ) 37.5 40.0 42.5 45.0 47.5 50.0 52.5 Accuracy (%) CIFAR-100-LT BCL Sweep (BSM, =2.0) =0.9 Macro =0.9 Few…
Figure 5
Figure 5. Figure 5: Effect of the BCL weight λBCL on CIFAR-100-LT. BCL is used as the base expert objective within CLEAR. clipping schedule is mild, the experts tend to have similar reliability, and both Macro Accuracy and Few-shot Accuracy remain relatively stable across different values…
Figure 6
Figure 6. Figure 6: Effect of post-hoc logit adjustment αLA. LA is applied as an optional inference-time correction within CLEAR. can improve tail-class expert representations, but should be regarded as an interchangeable expert-training objective within CLEAR [PITH_FULL_IMAGE:figures/fu…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 25 canonical work pages

  1. [1]

    Learning imbalanced datasets with label-distribution-aware margin loss

    Kaidi Cao, Colin Wei, Adrien Gaidon, Nikos Arechiga, and Tengyu Ma. Learning imbalanced datasets with label-distribution-aware margin loss. InAdvances in Neural Information Processing Systems (NeurIPS), 2019

    work page 2019

  2. [2]

    Parametric contrastive learning

    Jiequan Cui, Zhisheng Zhong, Shu Liu, Bei Yu, and Jiaya Jia. Parametric contrastive learning. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 715–724, 2021

    work page 2021

  3. [3]

    Generalized parametric contrastive learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    Jiequan Cui, Zhisheng Zhong, Zhuotao Tian, Shu Liu, Bei Yu, and Jiaya Jia. Generalized parametric contrastive learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. doi: 10.1109/TPAMI.2023.3278694

    arXiv 2024

  4. [4]

    Class-balanced loss based on effective number of samples

    Yin Cui, Menglin Jia, Tsung-Yi Lin, Yang Song, and Serge Belongie. Class-balanced loss based on effective number of samples. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9268–9277, 2019

    work page 2019

  5. [5]

    Disentan- gling label distribution for long-tailed visual recognition

    Youngkyu Hong, Seungju Han, Kwanghee Choi, Seokjun Seo, Beomsu Kim, and Buru Chang. Disentan- gling label distribution for long-tailed visual recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 6626–6636, 2021

    work page 2021

  6. [6]

    Ensemble partition sampling (eps) for improved multi-class classification.IEEE Access, 11:48221–48235, 2023

    Brahim Jabir, Isabel De La Torre Díez, Ernesto Francisco Bautista Thompson, Debora Libertad Ramírez Vargas, and Ángel Gabriel Kuc Castilla. Ensemble partition sampling (eps) for improved multi-class classification.IEEE Access, 11:48221–48235, 2023. doi: 10.1109/ACCESS.2023.3273925

    arXiv 2023

  7. [7]

    Decoupling representation and classifier for long-tailed recognition

    Bingyi Kang, Saining Xie, Marcus Rohrbach, Zhicheng Yan, Albert Gordo, Jiashi Feng, and Yan- nis Kalantidis. Decoupling representation and classifier for long-tailed recognition. InInternational Conference on Learning Representations (ICLR), 2020

    work page 2020

  8. [8]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. Focal loss for dense object detection. InProceedings of the IEEE International Conference on Computer Vision (ICCV), pages 2980–2988, 2017. 17

    work page 2017

Show all 28 references
  1. [9]

    Long-tailed classification with multi-granularity semantics

    Yuting Liu, Liu Yang, and Yu Wang. Long-tailed classification with multi-granularity semantics. In 2025 IEEE/CVF International Conference on Computer Vision (ICCV), pages 4285–4294, 2025. doi: 10.1109/ICCV51701.2025.00408

    2025

  2. [10]

    Large-scale long-tailed recognition in an open world

    Ziwei Liu, Zhongqi Miao, Xiaohang Zhan, Jiayun Wang, Boqing Gong, and Stella X Yu. Large-scale long-tailed recognition in an open world. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 2537–2546, 2019

    2019

  3. [11]

    Long-tailed recognition with model rebalancing

    Jiaan Luo, Feng Hong, Qiang Hu, Xiaofeng Cao, Feng Liu, and Jiangchao Yao. Long-tailed recognition with model rebalancing. InAdvances in Neural Information Processing Systems (NeurIPS), 2025

    2025

  4. [12]

    Aligned contrastive loss for long-tailed recognition.arXiv preprint arXiv:2506.01071, 2025

    Jiali Ma, Jiequan Cui, Kazuki Maeno, Lakshmi Subramanian, Jayashree Karlekar, Sugiri Pranata, and Hanwang Zhang. Aligned contrastive loss for long-tailed recognition.arXiv preprint arXiv:2506.01071, 2025

    2025 arXiv

  5. [13]

    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. InInternational Conference on Learning Representations, 2021

    2021

  6. [14]

    Enhanced long-tailed recognition with contrastive cutmix augmentation.arXiv preprint arXiv:2407.04911, 2024

    Haolin Pan, Yong Guo, Mianjie Yu, and Jian Chen. Enhanced long-tailed recognition with contrastive cutmix augmentation.arXiv preprint arXiv:2407.04911, 2024

    2024 arXiv

  7. [15]

    Balanced meta-softmax for long-tailed visual recognition

    Jiawei Ren, Cunjun Yu, Shunan Sheng, Xiao Ma, Haiyu Zhao, Shuai Yi, and Hongsheng Li. Balanced meta-softmax for long-tailed visual recognition. InAdvances in Neural Information Processing Systems (NeurIPS), 2020

    2020

  8. [16]

    In defense of One-Vs-All classification.Journal of Machine Learning Research, 5:101–141, 2004

    Ryan Rifkin and Aldebaro Klautau. In defense of One-Vs-All classification.Journal of Machine Learning Research, 5:101–141, 2004

    2004

  9. [17]

    Salehi and M

    A. Salehi and M. Khedmati. Hybrid clustering strategies for effective oversampling and undersampling in multiclass classification.Scientific Reports, 15:3460, 2025. doi: 10.1038/s41598-024-84786-2

    2025 doi

  10. [18]

    Distributional robustness loss for long-tail learning

    Dvir Samuel and Gal Chechik. Distributional robustness loss for long-tail learning. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021

    2021

  11. [19]

    Long-tailed classification by keeping the good and removing the bad momentum causal effect

    Kaihua Tang, Jianqiang Huang, and Hanwang Zhang. Long-tailed classification by keeping the good and removing the bad momentum causal effect. InAdvances in Neural Information Processing Systems (NeurIPS), 2020

    2020

  12. [20]

    Xudong Wang, Long Lian, Zhongqi Miao, Ziwei Liu, and Stella X. Yu. Long-tailed recognition by routing diverse distribution-aware experts. InInternational Conference on Learning Representations (ICLR), 2021

    2021

  13. [21]

    Balanced knowledge distillation for long-tailed learning.Neurocomputing, 527:36–46, 2023

    Shaoyu Zhang, Chen Chen, Xiyuan Hu, and Silong Peng. Balanced knowledge distillation for long-tailed learning.Neurocomputing, 527:36–46, 2023. doi: 10.1016/j.neucom.2023.01.063

    2023 doi

  14. [22]

    Distribution alignment: A unified framework for long-tail visual recognition

    Songyang Zhang, Zeming Li, Shipeng Yan, Xuming He, and Jian Sun. Distribution alignment: A unified framework for long-tail visual recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021

    2021

  15. [23]

    Self-supervised aggregation of diverse experts for test-agnostic long-tailed recognition

    Yifan Zhang, Bryan Hooi, Lanqing Hong, and Jiashi Feng. Self-supervised aggregation of diverse experts for test-agnostic long-tailed recognition. InAdvances in Neural Information Processing Systems (NeurIPS), volume 35, pages 34077–34090, 2022

    2022

  16. [24]

    MDCS: More diverse experts with consistency self-distillation for long-tailed recognition

    Qihao Zhao, Chen Jiang, Wei Hu, Fan Zhang, and Jun Liu. MDCS: More diverse experts with consistency self-distillation for long-tailed recognition. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 11597–11608, 2023

    2023

  17. [25]

    LTRL: Boosting long-tail recognition via reflective learning

    Qihao Zhao, Yalun Dai, Shen Lin, Wei Hu, Fan Zhang, and Jun Liu. LTRL: Boosting long-tail recognition via reflective learning. InEuropean Conference on Computer Vision (ECCV), 2024. 18

    2024

  18. [26]

    Improving calibration for long-tailed recognition

    Zhisheng Zhong, Jiequan Cui, Shu Liu, and Jiaya Jia. Improving calibration for long-tailed recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 16489–16498, 2021

    2021

  19. [27]

    BBN: Bilateral-branch network with cumulative learning for long-tailed visual recognition

    Boyan Zhou, Quan Cui, Xiu-Shen Wei, and Zhao-Min Chen. BBN: Bilateral-branch network with cumulative learning for long-tailed visual recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9716–9725, 2020

    2020

  20. [28]

    Balanced contrastive learning for long-tailed visual recognition

    Jianggang Zhu, Zheng Wang, Jingjing Chen, Yi-Ping Phoebe Chen, and Yu-Gang Jiang. Balanced contrastive learning for long-tailed visual recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 6908–6917, 2022. 19

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.