Pith. sign in

REVIEW 5 major objections 7 minor 54 references

Few-Shot Inspired Generative Zero-Shot Learning

T0 review · 5 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Generative zero-shot classification can match strong baselines while producing only a handful of synthetic prototypes per unseen class, because a few attribute-adjusted group centers carry the intra-class structure.

desk verdict A plausible efficiency-oriented take on generative ZSL that deserves a real referee, but the headline accuracy claim is not yet supported because the comparison runs on a different visual backbone. read the letter →

arxiv 2507.01026 v1 pith:PNBXQEFR submitted 2025-06-18 cs.LG

classification cs.LG
keywords zero-shotlearninggeneralizedfew-shot-inspiredattributescoringfeaturesynthesisprototypegenerationcontrastivesemanticregularization
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

This paper tries to establish that generative zero-shot learning does not need large-scale feature synthesis. It proposes FSIGenZ, which treats the task as closer to few-shot learning: for each unseen class it builds only a small set of group-level prototypes, computed from seen-class centers via sparse coding, after re-scoring class attributes to approximate instance-level variability. A semantic-aware contrastive classifier trained on these prototypes plus real seen features, with a semantic regularization loss, compensates for the resulting class imbalance. The payoff would be a zero-shot pipeline with much lower compute and data, closer to the original spirit of zero-shot learning, at accuracy comparable to state-of-the-art generative methods on SUN, AwA2, and CUB.

What carries the argument

The load-bearing object is the group-level prototype: a synthetic center $\mu^u_{kp}=M_s\alpha_p$, built by applying sparse-code coefficients $\alpha_p$ (which encode each unseen class's attribute vector onto seen class attributes, Eq. 4) to the matrix of seen-class visual centers $M_s$, with the regularization strength $\lambda$ varied to produce several subgroup prototypes per class. MSAS (Eqs. 1-2) re-scores the original class attributes by reweighting and thresholding to approximate instance-level variability. The classifier is a semantic-aware contrastive network that fuses visual features with class-semantic embeddings and is trained with a DPSR loss that injects class-to-class semantic similarities into unseen-class supervision. These pieces replace large-scale feature generation with a small set of informative training points.

What would settle it

Using the paper's own t-SNE setup, compute the distance between each estimated prototype $\mu^u_{kp}$ and the nearest true k-means sub-cluster center of real unseen features, normalized by the sub-cluster's standard deviation; if that normalized distance is large, or if removing any single prototype changes accuracy materially, the claim that a handful of prototypes represents the unseen distribution fails.

Watch

Extended reading notes

Core claim

The central discovery is that the usual generative zero-shot recipe—synthesize tens of thousands of visual features for unseen classes and then train a classifier on them—overshoots. FSIGenZ shows that the unseen-class feature distribution can be represented by a compact set of prototypes, one per semantic subgroup, derived from seen-class centers by transferring the relation between attribute embeddings and visual features. Instance-level attribute variability, which class-level attribute vectors ignore, is approximated by rescaling attributes through MSAS and by drawing multiple sparse-code solutions with different regularization strengths $\lambda$. The trained classifier, regularized by DPSR, classifies unseen classes at levels comparable to methods that synthesize thousands of features per class. On the three benchmarks, the reported totals are 1080, 900, and 500 synthetic features for SUN, AwA2, and CUB, with harmonic-mean GZSL scores of 45.9%, 74.2%, and 69.1%.

Load-bearing premise

The method assumes that a few prototype points—linear combinations of seen-class visual centers with sparse coefficients derived from attributes—capture enough of each unseen class's true feature distribution for the classifier to generalize.

Editorial extensions

If this is right

  • On the three benchmarks, matching or exceeding generative baselines requires only 1,080 (SUN), 900 (AwA2), and 500 (CUB) total synthetic features, compared with tens of thousands in prior generative methods.
  • Feature generation no longer needs adversarial training: prototypes come from sparse coding and class-center combinations, so the computational cost of GAN/VAE training is removed from the pipeline.
  • The ablation results show DPSR is responsible for the GZSL gains: removing it drops harmonic mean on AwA2 from 74.2% to 16.1%, so semantic regularization is what makes the low-data regime usable.
  • At inference the model uses raw contrastive scores with no regularization, so the reported performance is achieved with the same simple scoring rule used by embedding-based methods.
  • Estimated subgroup prototypes align with k-means sub-clusters of real images in the paper's t-SNE visualizations, supporting the claim that the prototypes capture genuine intra-class structure.

Reading between the lines

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

  • An extension the paper does not pursue is using the prototype count as a budget: one could tie the number of regularization strengths per class to attribute variance or class difficulty, and the paper's own sensitivity analysis (Fig. 5) suggests this choice materially affects accuracy.
  • A stress test for the transfer assumption would be to train the sparse-coding relation on one benchmark's seen classes and synthesize prototypes for another dataset's unseen classes; if alignment with true sub-clusters degrades, the bottleneck is the seen-to-unseen attribute relation.
  • The framing suggests a direct few-shot transfer path: treating the estimated prototypes as shots would let the same classifier be evaluated under standard K-shot protocols, which would quantify how many prototypes are genuinely needed.
  • Beyond datasets with dense expert attributes, one could test whether MSAS's thresholding still helps with noisy or machine-learned attribute scores; the gain may shrink when the original scores are already calibrated.
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

5 major / 7 minor

Summary. FSIGenZ proposes a generative zero-shot learning approach that synthesizes only a small set of group-level prototypes per unseen class, using Model-Specific Attribute Scoring (MSAS) to re-weight class attributes and a Dual-Purpose Semantic Regularization (DPSR) trained with a semantic-aware contrastive classifier (SCC). The method is evaluated on SUN, AwA2, and CUB, reporting accuracy comparable to state-of-the-art generative models while using orders of magnitude fewer synthetic features (Table 2).

Significance. If the claims are substantiated, FSIGenZ would be a computationally lightweight alternative to GAN/VAE-based feature synthesis, preserving the ZSL assumption of no unseen-class supervision during training. The paper's strengths are the clear problem framing (few-shot-inspired), the fully non-adversarial synthesis pipeline, and the detailed ablations showing the contribution of each component. However, the central empirical claim is currently confounded by a feature-backbone mismatch and a lack of reproducibility details, so the value of the method over its backbone choice is not yet demonstrated.

major comments (5)
  1. [Section 4.1, Tables 1-2] FSIGenZ uses 786-dimensional ViT-Base features (Section 4.1), while most baselines in Table 1 (e.g., f-CLSWGAN, CE-GZSL, FREE, DPCN) use ResNet-101 features. Because ViT-Base features are known to be substantially stronger for zero-shot classification, the reported T1 and H comparisons are not apples-to-apples, and the central claim of 'competitive with far fewer synthetic features' (Section 4.2) may be attributable to the backbone rather than to MSAS, DPSR, or the prototype synthesis. The authors must provide matched-backbone experiments, such as re-running baselines on ViT-Base features or running FSIGenZ on ResNet-101 features, and report those results.
  2. [Section 3.4, Eq. (4), Section 4.1] The claim that tuning λ yields 'varied, sparse representations' (Section 3.4) is not supported: Eq. (4) uses an L2 penalty on α, which does not induce sparsity, and λ is sampled uniformly from [1, 1.02] (Section 4.1), producing near-identical α vectors and thus near-identical prototypes μ^u_k from Eq. (5). The paper should either use an L1 sparsity penalty, allow λ to vary over a wider range, or demonstrate quantitatively (e.g., pairwise prototype distances or downstream accuracy as a function of λ) that the generated prototypes are actually diverse and that this diversity drives the performance.
  3. [Section 3.5, Eq. (14)] The normalization step in Eq. (14) uses the undefined symbol \bar{s}_pq; without a definition, the DPSR loss in Eq. (11) is incompletely specified. Additionally, the text states that φ 'discourages any single similarity score, particularly self-similarity, from dominating,' but the L2 penalty in Eq. (13) does not specifically penalize self-similarity; the intended mechanism should be clarified.
  4. [Section 4.1 and 4.3] Hyperparameters W_A, T_h, β, λ, and the number of synthetic features per class are set per dataset (Section 4.1) without a described validation protocol. Since unseen classes are not available during training in ZSL, tuning on the test split is a risk; the authors must specify how hyperparameters were selected (e.g., a held-out validation set of unseen classes or cross-validation on seen classes) and report results over multiple seeds with error bars.
  5. [Section 3.4, Figure 7] The assumption that the relation function R_s transfers from seen to unseen classes (Eq. 3-5) is the core of the synthesis method, yet it is validated only through a qualitative t-SNE visualization of 'seven randomly selected' clusters (Figure 7). A quantitative evaluation (e.g., distance between estimated prototypes and true cluster centers, or classification accuracy using only synthetic prototypes) is needed to show that the prototypes are informative.
minor comments (7)
  1. [Algorithm 1, line 5] Algorithm 1 line 5 states 'Apply MSAS on class attributes: A ← Ao', which omits the thresholding and scaling described in Eqs. (1)-(2); the algorithm should be updated to reflect the MSAS computation.
  2. [Section 3.4] Eq. (4) is described as 'Sparse Coding' but uses an L2 penalty; the terminology is misleading since L2 regularization does not produce sparse coefficients.
  3. [Section 4.2] The claim that FSIGenZ's performance is 'unmatched by any other method in the table' is an overstatement: on CUB, DPCN achieves T1=80.1% vs FSIGenZ's 73.0%, and on SUN, LCR-GAN achieves H=49.8% vs FSIGenZ's 45.9%.
  4. [Table 1 caption] The caption's highlighting scheme (bold, underline, red, blue) is not fully explained: please clarify which comparisons the red/blue highlights refer to and how ties are handled.
  5. [Figure 7] The t-SNE visualization does not report the dimensionality reduction hyperparameters (e.g., perplexity, number of iterations) or the number of real clusters used, making it difficult to assess the alignment quantitatively.
  6. [Section 4.1] The statement that 'λ takes the same number of random values between 1 and 1.02' is ambiguous: it is unclear whether these values are drawn once per dataset or per run, and whether the number of values per class is considered a hyperparameter that should be tuned.
  7. [Section 3.4] The description of MDP [30] as inspiration is vague; please clarify how the proposed multi-source knowledge transfer relates to MDP and what specific elements are adopted.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the unseen-class prototypes are extrapolated from seen-class centers via sparse attribute coding, and the only author self-citation is used for standard taxonomy, not as load-bearing evidence.

full rationale

FSIGenZ's derivation chain is self-contained. Unseen prototypes are computed as linear combinations of seen class feature centers with sparse coefficients obtained by coding unseen attribute vectors against seen attribute vectors (Eqs. 4-5); no unseen labels or test features enter the training objective (Eqs. 9-12). MSAS is an attribute reweighting with hyperparameters, and DPSR is semantic label smoothing built from the same class-attribute matrix used to define the problem; both are design choices rather than fitted predictions of unseen labels. The paper's only self-citation is [3], a GZSL review by two of the authors, cited for the standard CZSL/GZSL taxonomy and survey background; it does not support the prototype-generation or contrastive-classification claims, so it is not load-bearing. The remaining concerns, namely the ViT-Base versus ResNet-101 backbone mismatch in Table 1 and per-dataset hyperparameter selection without a stated validation protocol, are threats to the comparative accuracy claim rather than circular reductions.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

No new physical or external entities are introduced. The 'prototypes', 'MSAS-adjusted scores', and 'DPSR similarities' are constructs internal to the method, and the paper provides no independent falsifiable handle for them beyond the reported benchmark results.

free parameters (5)
  • W_A (MSAS weight) = 0.005 (SUN), 0.08 (AwA2), 0.3 (CUB)
    Scales attribute scores in Eq. 1; tuned per dataset in Section 4.1.
  • T_h (MSAS threshold) = 0.7 (SUN), 0.8 (AwA2), 0.7 (CUB)
    Threshold in Eq. 2 that masks low attributes; tuned per dataset.
  • λ (sparse coding regularization) = random values in [1,1.02]; count per class 15/90/10 for SUN/AwA2/CUB
    Controls number of prototypes and their variation in Eq. 4/5; values are random and not specified.
  • β (loss weight) = 0.2
    Weights unseen-class loss in Eq. 12; chosen from sensitivity analysis (Fig. 6).
  • φ (DPSR regularization) = not specified
    Regularization in Eq. 13 prevents trivial similarity solutions; value not given in paper.
assumptions (5)
  • domain assumption Class-level attributes are predictive of visual features, so sparse codes of unseen attributes onto seen attributes yield valid visual prototypes (Eq. 4, 5).
    Core ZSL transfer assumption, inherited from MDP [30].
  • ad hoc to paper Instance-level attribute variability can be approximated by a global thresholding and scaling of class-level attributes (Eq. 1-2).
    MSAS assumes a simple piecewise linear transform captures missing or partial attributes; supported only by qualitative Fig. 2.
  • domain assumption Semantic similarity between classes (from attributes) is a good guide for softening classifier outputs (Section 3.5, Eq. 13).
    DPSR relies on attribute-based class relationships transferring to visual decision boundaries.
  • ad hoc to paper A small set of group prototypes (e.g., 10-90 per class) is sufficient training signal for unseen classes.
    The few-shot-inspired premise; supported only by the paper's own experiments, not by an external benchmark.
  • standard math Sparse coding solution to Eq. 4 is well-posed under the given λ and data.
    Standard optimization assumption; convex when λ>0.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Few-Shot Inspired Generative Zero-Shot Learning." pith.science (2026). https://pith.science/paper/PNBXQEFR

@misc{pith2026250701026,
  author       = {Pith},
  title        = {Pith review of: Few-Shot Inspired Generative Zero-Shot Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PNBXQEFR}},
  note         = {Machine review of arXiv:2507.01026}
}
read the original abstract

Generative zero-shot learning (ZSL) methods typically synthesize visual features for unseen classes using predefined semantic attributes, followed by training a fully supervised classification model. While effective, these methods require substantial computational resources and extensive synthetic data, thereby relaxing the original ZSL assumptions. In this paper, we propose FSIGenZ, a few-shot-inspired generative ZSL framework that reduces reliance on large-scale feature synthesis. Our key insight is that class-level attributes exhibit instance-level variability, i.e., some attributes may be absent or partially visible, yet conventional ZSL methods treat them as uniformly present. To address this, we introduce Model-Specific Attribute Scoring (MSAS), which dynamically re-scores class attributes based on model-specific optimization to approximate instance-level variability without access to unseen data. We further estimate group-level prototypes as clusters of instances based on MSAS-adjusted attribute scores, which serve as representative synthetic features for each unseen class. To mitigate the resulting data imbalance, we introduce a Dual-Purpose Semantic Regularization (DPSR) strategy while training a semantic-aware contrastive classifier (SCC) using these prototypes. Experiments on SUN, AwA2, and CUB benchmarks demonstrate that FSIGenZ achieves competitive performance using far fewer synthetic features.

Figures

Figures reproduced from arXiv: 2507.01026 by the authors.

Figure 1
Figure 1. Conceptual illustration: FSL is a less relaxed problem formulation of ZSL than the supervised formulation. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Our observations: Even for four attributes—Furry, Hooves, Tail, and Quadrupedal—(a) shows ungrouped [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overview of the FSIGenZ framework. The model comprises two main phases: feature generation (above [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: DPSR conceptualized in a flowchart: regularization of model’s transferable and discriminative confidence to [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Results concerning varying synthetic instances for each unseen class of the (a) SUN, (b) CUB, and (c) AwA2 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Results for various β values on (a) SUN, (b) CUB, and (c) AwA2 datasets [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: t-SNE visualization of seven randomly selected real (circle) and estimated (cross) clusters of the (a) SUN, (b) [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 50 canonical work pages

  1. [1]

    Zero-shot learning—a comprehensive evaluation of the good, the bad and the ugly.IEEE Trans

    Yongqin Xian, Christoph H Lampert, Bernt Schiele, and Zeynep Akata. Zero-shot learning—a comprehensive evaluation of the good, the bad and the ugly.IEEE Trans. Pattern Anal. Mach. Intell., 41(9):2251–2265, 2018

  2. [2]

    Attribute-based classification for zero-shot visual object categorization.IEEE Trans

    Christoph H Lampert, Hannes Nickisch, and Stefan Harmeling. Attribute-based classification for zero-shot visual object categorization.IEEE Trans. Pattern Anal. Mach. Intell., 36(3):453–465, 2013

  3. [3]

    A review of generalized zero-shot learning methods.IEEE Trans

    Farhad Pourpanah, Moloud Abdar, Yuxuan Luo, Xinlei Zhou, Ran Wang, Chee Peng Lim, Xi-Zhao Wang, and QM Jonathan Wu. A review of generalized zero-shot learning methods.IEEE Trans. Pattern Anal. Mach. Intell., 45(4):4051–4070, 2022

  4. [4]

    Duet: Cross-modal semantic grounding for contrastive zero-shot learning

    Zhuo Chen, Yufeng Huang, Jiaoyan Chen, Yuxia Geng, Wen Zhang, Yin Fang, Jeff Z Pan, and Huajun Chen. Duet: Cross-modal semantic grounding for contrastive zero-shot learning. InProc. AAAI Conf. Artif. Intell., volume 37, pages 405–413, 2023

  5. [5]

    Generative adversarial networks.Commun

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks.Commun. ACM, 63(11):139–144, 2020

  6. [6]

    Auto-encoding variational bayes.arXiv preprint arXiv:1312.6114, 2013

    Diederik P Kingma and Max Welling. Auto-encoding variational bayes.arXiv preprint arXiv:1312.6114, 2013

  7. [7]

    Feature generating networks for zero-shot learning

    Yongqin Xian, Tobias Lorenz, Bernt Schiele, and Zeynep Akata. Feature generating networks for zero-shot learning. InProc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 5542–5551, 2018

  8. [8]

    Synthetic sample selection for generalized zero-shot learning

    Shreyank N Gowda. Synthetic sample selection for generalized zero-shot learning. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 58–67, 2023

Show all 54 references
  1. [9]

    Sun attribute database: Discovering, annotating, and recognizing scene attributes

    Genevieve Patterson and James Hays. Sun attribute database: Discovering, annotating, and recognizing scene attributes. InProc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 2751–2758. IEEE, 2012

  2. [10]

    The caltech-ucsd birds-200- 2011 dataset, 2011

    Catherine Wah, Steve Branson, Peter Welinder, Pietro Perona, and Serge Belongie. The caltech-ucsd birds-200- 2011 dataset, 2011

  3. [11]

    Preserving semantic relations for zero-shot learning

    Yashas Annadani and Soma Biswas. Preserving semantic relations for zero-shot learning. InProc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 7603–7612, 2018

  4. [12]

    Explainable zero-shot learning via attentive graph convolutional network and knowledge graphs.Semant

    Yuxia Geng, Jiaoyan Chen, Zhiquan Ye, Zonggang Yuan, Wei Zhang, and Huajun Chen. Explainable zero-shot learning via attentive graph convolutional network and knowledge graphs.Semant. Web, 12(5):741–765, 2021

  5. [13]

    Label-activating framework for zero-shot learning.Neural Netw., 121:1–9, 2020

    Yang Liu, Xinbo Gao, Quanxue Gao, Jungong Han, and Ling Shao. Label-activating framework for zero-shot learning.Neural Netw., 121:1–9, 2020

  6. [14]

    Co-representation network for generalized zero-shot learning

    Fei Zhang and Guangming Shi. Co-representation network for generalized zero-shot learning. InProc. Int. Conf. Mach. Learn., pages 7434–7443. PMLR, 2019

  7. [15]

    Zero-shot learning via semantic similarity embedding

    Ziming Zhang and Venkatesh Saligrama. Zero-shot learning via semantic similarity embedding. InProc. IEEE/CVF Int. Conf. Comput. Vis., pages 4166–4174, 2015. 11

  8. [16]

    f-vaegan-d2: A feature generating framework for any-shot learning

    Yongqin Xian, Saurabh Sharma, Bernt Schiele, and Zeynep Akata. f-vaegan-d2: A feature generating framework for any-shot learning. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 10275–10284, 2019

  9. [17]

    Free: Feature refinement for generalized zero-shot learning

    Shiming Chen, Wenjie Wang, Beihao Xia, Qinmu Peng, Xinge You, Feng Zheng, and Ling Shao. Free: Feature refinement for generalized zero-shot learning. InProc. IEEE/CVF Int. Conf. Comput. Vis., pages 122–131, 2021

  10. [18]

    Zerogen: Efficient zero-shot learning via dataset generation.arXiv preprint arXiv:2202.07922, 2022

    Jiacheng Ye, Jiahui Gao, Qintong Li, Hang Xu, Jiangtao Feng, Zhiyong Wu, Tao Yu, and Lingpeng Kong. Zerogen: Efficient zero-shot learning via dataset generation.arXiv preprint arXiv:2202.07922, 2022

  11. [19]

    Contrastive embedding for generalized zero-shot learning

    Zongyan Han, Zhenyong Fu, Shuo Chen, and Jian Yang. Contrastive embedding for generalized zero-shot learning. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 2371–2381, 2021

  12. [20]

    Re-gzsl: Relation extrapolation for generalized zero-shot learning

    Yao Wu, Xia Kong, Yuan Xie, and Yanyun Qu. Re-gzsl: Relation extrapolation for generalized zero-shot learning. IEEE Trans. Circuits Syst. Video Technol., 2024

  13. [21]

    En- compactness: Self-distillation embedding & contrastive generation for generalized zero-shot learning

    Xia Kong, Zuodong Gao, Xiaofan Li, Ming Hong, Jun Liu, Chengjie Wang, Yuan Xie, and Yanyun Qu. En- compactness: Self-distillation embedding & contrastive generation for generalized zero-shot learning. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 9306–9315, 2022

  14. [22]

    Deep multimodal representation learning: A survey.Ieee Access, 7:63373–63394, 2019

    Wenzhong Guo, Jianwen Wang, and Shiping Wang. Deep multimodal representation learning: A survey.Ieee Access, 7:63373–63394, 2019

  15. [23]

    steerability

    Ali Jahanian, Lucy Chai, and Phillip Isola. On the" steerability" of generative adversarial networks.arXiv preprint arXiv:1907.07171, 2019

  16. [24]

    Non-generative generalized zero-shot learning via task-correlated disentanglement and controllable samples synthesis

    Yaogong Feng, Xiaowen Huang, Pengbo Yang, Jian Yu, and Jitao Sang. Non-generative generalized zero-shot learning via task-correlated disentanglement and controllable samples synthesis. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 9346–9355, 2022

  17. [25]

    No adversaries to zero-shot learning: Distilling an ensemble of gaussian feature generators.IEEE Trans

    Jacopo Cavazza, Vittorio Murino, and Alessio Del Bue. No adversaries to zero-shot learning: Distilling an ensemble of gaussian feature generators.IEEE Trans. Pattern Anal. Mach. Intell., 2023

  18. [26]

    Compositional zero-shot learning via fine-grained dense feature composition

    Dat Huynh and Ehsan Elhamifar. Compositional zero-shot learning via fine-grained dense feature composition. In Adv. Neural Inf. Process. Syst., volume 33, pages 19849–19860, 2020

  19. [27]

    Attribute-based synthetic network (abs-net): Learning more from pseudo feature representations.Pattern Recognit., 80:129–142, 2018

    Jiang Lu, Jin Li, Ziang Yan, Fenghua Mei, and Changshui Zhang. Attribute-based synthetic network (abs-net): Learning more from pseudo feature representations.Pattern Recognit., 80:129–142, 2018

  20. [28]

    Zero and few shot learning with semantic feature synthesis and competitive learning.IEEE Trans

    Jiechao Guan, Zhiwu Lu, Tao Xiang, Aoxue Li, An Zhao, and Ji-Rong Wen. Zero and few shot learning with semantic feature synthesis and competitive learning.IEEE Trans. Pattern Anal. Mach. Intell., 43(7):2510–2523, 2020

  21. [29]

    Adaptive and generative zero-shot learning

    Yu-Ying Chou, Hsuan-Tien Lin, and Tyng-Luh Liu. Adaptive and generative zero-shot learning. InInt. Conf. Learn. Represent, 2021

  22. [30]

    Zero-shot learning posed as a missing data problem

    Bo Zhao, Botong Wu, Tianfu Wu, and Yizhou Wang. Zero-shot learning posed as a missing data problem. In Proc. IEEE Int. Conf. Comput. Vis. Workshops, pages 2616–2622, 2017

  23. [31]

    Fine-grained generalized zero-shot learning via dense attribute-based attention

    Dat Huynh and Ehsan Elhamifar. Fine-grained generalized zero-shot learning via dense attribute-based attention. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 4483–4493, 2020

  24. [32]

    Attribute prototype network for zero-shot learning

    Wenjia Xu, Yongqin Xian, Jiuniu Wang, Bernt Schiele, and Zeynep Akata. Attribute prototype network for zero-shot learning. InAdv. Neural Inf. Process. Syst., volume 33, pages 21969–21980, 2020

  25. [33]

    Semantic-guided multi-attention localization for zero-shot learning

    Yizhe Zhu, Jianwen Xie, Zhiqiang Tang, Xi Peng, and Ahmed Elgammal. Semantic-guided multi-attention localization for zero-shot learning. InAdv. Neural Inf. Process. Syst., volume 32, 2019

  26. [34]

    Transferable contrastive network for generalized zero-shot learning

    Huajie Jiang, Ruiping Wang, Shiguang Shan, and Xilin Chen. Transferable contrastive network for generalized zero-shot learning. InProc. IEEE/CVF Int. Conf. Comput. Vis., pages 9765–9774, 2019

  27. [35]

    Multi-head self-attention via vision transformer for zero-shot learning.arXiv preprint arXiv:2108.00045, 2021

    Faisal Alamri and Anjan Dutta. Multi-head self-attention via vision transformer for zero-shot learning.arXiv preprint arXiv:2108.00045, 2021

  28. [36]

    Msdn: Mutually semantic distillation network for zero-shot learning

    Shiming Chen, Ziming Hong, Guo-Sen Xie, Wenhan Yang, Qinmu Peng, Kai Wang, Jian Zhao, and Xinge You. Msdn: Mutually semantic distillation network for zero-shot learning. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 7612–7621, 2022

  29. [37]

    Semantic-guided class-imbalance learning model for zero-shot image classification.IEEE Trans

    Zhong Ji, Xuejie Yu, Yunlong Yu, Yanwei Pang, and Zhongfei Zhang. Semantic-guided class-imbalance learning model for zero-shot image classification.IEEE Trans. Cybern., 52(7):6543–6554, 2022

  30. [38]

    Diversity-boosted generalization-specialization balancing for zero-shot learning.IEEE Trans

    Yun Li, Zhe Liu, Xiaojun Chang, Julian McAuley, and Lina Yao. Diversity-boosted generalization-specialization balancing for zero-shot learning.IEEE Trans. Multimedia, 25:8372–8382, 2023. 12

  31. [39]

    Prototype rectification for zero-shot learning.Pattern Recognit., 156:110750, 2024

    Yuanyuan Yi, Guolei Zeng, Bocheng Ren, Laurence T Yang, Bin Chai, and Yuxin Li. Prototype rectification for zero-shot learning.Pattern Recognit., 156:110750, 2024

  32. [40]

    Zs-vat: Learning unbiased attribute knowledge for zero-shot recognition through visual attribute transformer.IEEE Trans

    Zongyan Han, Zhenyong Fu, Shuo Chen, Le Hui, Guangyu Li, Jian Yang, and Chang Wen Chen. Zs-vat: Learning unbiased attribute knowledge for zero-shot recognition through visual attribute transformer.IEEE Trans. Neural Netw. Learn. Syst., 36(4):7025–7036, 2025

  33. [41]

    Generalized zero-shot learning via over-complete distribution

    Rohit Keshari, Richa Singh, and Mayank Vatsa. Generalized zero-shot learning via over-complete distribution. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 13300–13308, 2020

  34. [42]

    Latent embedding feedback and discriminative features for zero-shot classification

    Sanath Narayan, Akshita Gupta, Fahad Shahbaz Khan, Cees GM Snoek, and Ling Shao. Latent embedding feedback and discriminative features for zero-shot classification. InProc. Eur. Conf. Comput. Vis, pages 479–495. Springer, 2020

  35. [43]

    Hsva: Hierarchical semantic-visual adaptation for zero-shot learning

    Shiming Chen, Guosen Xie, Yang Liu, Qinmu Peng, Baigui Sun, Hao Li, Xinge You, and Ling Shao. Hsva: Hierarchical semantic-visual adaptation for zero-shot learning. InAdv. Neural Inf. Process. Syst., volume 34, pages 16622–16634, 2021

  36. [44]

    Task aligned generative meta-learning for zero-shot learning

    Zhe Liu, Yun Li, Lina Yao, Xianzhi Wang, and Guodong Long. Task aligned generative meta-learning for zero-shot learning. InProc. AAAI Conf. Artif. Intell., volume 35, pages 8723–8731, 2021

  37. [45]

    Counterfactual zero-shot and open-set visual recognition

    Zhongqi Yue, Tan Wang, Qianru Sun, Xian-Sheng Hua, and Hanwang Zhang. Counterfactual zero-shot and open-set visual recognition. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pages 15404–15414, 2021

  38. [46]

    Semantic feature extraction for generalized zero-shot learning

    Junhan Kim, Kyuhong Shim, and Byonghyo Shim. Semantic feature extraction for generalized zero-shot learning. InProc. AAAI Conf. Artif. Intell., volume 36, pages 1166–1173, 2022

  39. [47]

    Learning mlatent representations for generalized zero-shot learning.IEEE Trans

    Yalan Ye, Tongjie Pan, Tonghoujun Luo, Jingjing Li, and Heng Tao Shen. Learning mlatent representations for generalized zero-shot learning.IEEE Trans. Multimedia, 25:2252–2265, 2023

  40. [48]

    Dual-aligned feature confusion alleviation for generalized zero-shot learning.IEEE Trans

    Hongzu Su, Jingjing Li, Ke Lu, Lei Zhu, and Heng Tao Shen. Dual-aligned feature confusion alleviation for generalized zero-shot learning.IEEE Trans. Circuits Syst. Video Technol., 33(8):3774–3785, 2023

  41. [49]

    Zero-shot learning with attentive region embedding and enhanced semantics.IEEE Trans

    Yang Liu, Yuhao Dang, Xinbo Gao, Jungong Han, and Ling Shao. Zero-shot learning with attentive region embedding and enhanced semantics.IEEE Trans. Neural Netw. Learn. Syst., 35(3):4220–4231, 2024

  42. [50]

    Joint feature generation and open-set prototype learning for generalized zero-shot open-set classification.Pattern Recognit., 147:110133, 2024

    Xiao Li, Min Fang, and Zhibo Zhai. Joint feature generation and open-set prototype learning for generalized zero-shot open-set classification.Pattern Recognit., 147:110133, 2024

  43. [51]

    Towards discriminative feature generation for generalized zero-shot learning.IEEE Trans

    Jiannan Ge, Hongtao Xie, Pandeng Li, Lingxi Xie, Shaobo Min, and Yongdong Zhang. Towards discriminative feature generation for generalized zero-shot learning.IEEE Trans. Multimedia, 2024

  44. [52]

    Dual prototype contrastive network for generalized zero-shot learning.IEEE Trans

    Huajie Jiang, Zhengxian Li, Yongli Hu, Baocai Yin, Jian Yang, Anton van den Hengel, Ming-Hsuan Yang, and Yuankai Qi. Dual prototype contrastive network for generalized zero-shot learning.IEEE Trans. Circuits Syst. Video Technol., 35(2):1111–1122, 2025

  45. [53]

    Class-wise and instance-wise contrastive learning for zero-shot learning based on vaegan.Expert Syst

    Baolong Zheng, Zhanshan Li, and Jingyao Li. Class-wise and instance-wise contrastive learning for zero-shot learning based on vaegan.Expert Syst. Appl., page 126671, 2025

  46. [54]

    An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929, 2020

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at...

Pith tools

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