Pith. sign in

REVIEW 4 major objections 6 minor 58 references

Beyond Class Tokens: LLM-guided Dominant Property Mining for Few-shot Classification

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read LLM-picked properties push few-shot CLIP past class tokens

desk verdict A plausible new recipe for LLM-guided property tokens in CLIP few-shot classification; the core ablation holds, but missing numeric results, an unfair zero-shot table, and no retrieval ablation keep it conditional. read the letter →

arxiv 2507.20511 v2 pith:T3RCOM4I submitted 2025-07-28 cs.CV

Wei Zhuo , Runjie Luo , Wufeng Xue , Linlin Shen This is my paper
classification cs.CV
keywords few-shotlearningCLIPpropertytokenscontrastiveLLM-generateddescriptionscachemodelfine-grainedclassificationdomaingeneralization
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

Few-shot classification with CLIP normally compares an image's single class token against class-name text embeddings, and cache-model variants add only class-level support features. This paper argues that the class token loses localized visual detail and therefore cannot match fine-grained text descriptions, a failure it names fine-grained alignment deficiency. To repair this, BCT-CLIP adds multiple property tokens per image, generated by cross-attending learnable queries to image patches, and trains them with contrastive learning against a small set of 'dominating properties' selected from LLM-generated descriptions. The paper reports that adding property-cache scores to class-cache scores improves ImageNet 16-shot accuracy from 65.09% to 66.40%, and that the full method outperforms prior CLIP-based few-shot methods on most of 11 datasets and on out-of-distribution shifts.

What carries the argument

The Multi-Property Generator (MPG) is a small module of two cross-attention layers in which M learnable property-token queries attend to CLIP image patch features and are refined by a group-wise feed-forward network. The MPG is supervised by an LLM-based property retrieval pipeline that encodes LLM descriptions, K-means clusters them, ranks clusters by cosine similarity between support class tokens and cluster text embeddings using Eq. (4), and keeps the top-M clusters as one-vs-one positive annotations; the contrastive loss in Eq. (5) pulls each property token to its positive description and pushes it from hard and general negatives. Class and property prototypes from the support set form two cache models whose scores are combined with CLIP zero-shot scores in Eq. (6) and trained with cross-entropy in Eq. (7).

What would settle it

Retrain BCT-CLIP with the top-M clusters chosen by Eq. (4) replaced by random clusters of the same size; if ImageNet 16-shot accuracy stays within noise of 66.40%, the retrieval step is not responsible for the gain. A complementary test would corrupt local attributes in support images, such as recolored beaks, and check whether property tokens and selected descriptions track the alteration.

Watch

Extended reading notes

Core claim

The central claim is that the class token is not enough for few-shot visual-language classification and that class-specific localized properties recovered from LLM text can be converted into trainable visual property tokens. The paper claims its Multi-Property Generator can learn these tokens from as few as one to sixteen support images, and that its LLM retrieval procedure, by clustering descriptions and selecting clusters most similar to support images, identifies stable dominant properties that survive training. On the paper's terms, aligning property tokens to these selected descriptions through InfoNCE contrastive learning gives the model a structural image representation, and the hybrid cache built from class and property prototypes is what lifts classification accuracy on 11 benchmarks and on ImageNet-V2 and ImageNet-Sketch.

Load-bearing premise

The top-M cluster selection assumes that a support image's CLIP class token, which the paper argues is too coarse for fine-grained alignment, is still accurate enough to rank fine-grained property clusters by cosine similarity, and that the chosen clusters are faithful to the true class content.

Editorial extensions

If this is right

  • The hybrid cache score combining class and property caches outperforms either cache alone on ImageNet 16-shot (66.40% vs 65.09% class-cache-only), showing the property tokens add complementary information.
  • The property-cache-only score (65.60%) beats the class-cache-only score (65.09%), showing localized property tokens alone carry more discriminative signal than class tokens.
  • The method works across 11 datasets and on ImageNet-V2 and ImageNet-Sketch out-of-distribution sets, indicating the property-token representation generalizes beyond the training distribution.
  • Because property descriptions come from an LLM and are filtered by support-image similarity, the method does not need base-class training data and can be applied to novel classes with just a few shots.

Reading between the lines

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

  • The contribution of the top-M retrieval step could be isolated by replacing it with randomly chosen clusters of equal size; if accuracy stays high, the class-token ranking is not the source of the gain.
  • Because 'dominating properties' are clusters shared across classes, a natural extension is cross-dataset transfer, using property clusters learned on one dataset to seed another.
  • The main practical cost is LLM annotation, which could be amortized by precomputing descriptions per dataset once, since the MPG only adds 0.92 GFLOPS at test time.
  • Property tokens should be tested on fine-grained hard negatives within the same superclass to directly verify the proposed 'fine-grained alignment deficiency' fix.
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

4 major / 6 minor

Summary. The paper proposes BCT-CLIP, a few-shot classification method built on CLIP that augments the standard class token with multiple property tokens. A Multi-Property Generator (MPG) produces M patch-aware property tokens per image via cross-attention. LLM-generated textual descriptions are clustered, and the top-M clusters are selected through CLIP class-token similarity with support images. Property tokens are trained with an InfoNCE-style contrastive loss against the selected descriptions, and classification uses a hybrid cache model combining class-token and property-token caches. The authors report improvements over existing CLIP-based few-shot methods on ImageNet (Table I) and claim consistent gains across 11 datasets (Figs. 5 and 6), with ablations on ImageNet (Table III) and OOD robustness (Table II).

Significance. If the central claim is upheld, the idea of using LLM-generated and visually grounded 'dominant properties' to enrich CLIP representations for few-shot classification is a timely and potentially useful contribution. The paper presents a detailed pipeline, including the MPG architecture, the clustering-based property selection, the contrastive objective, and the hybrid cache model, and it provides an ablation showing that the property-cache score alone (65.60) exceeds the class-cache score (65.09) on ImageNet 16-shot, which is an interesting result. The complexity analysis and the study of learned alpha/beta weights are also welcome. However, the evidence currently has important gaps: the 11-dataset results are only shown as plots without variance, the Table IV comparison against zero-shot methods is not a fair zero-shot comparison for the proposed method, and there is no ablation isolating the contribution of the property-retrieval mechanism. These issues prevent the paper from fully supporting its strong causal claims at this stage.

major comments (4)
  1. [Section VII.C.f, Table IV] The comparison labeled 'zero-shot methods' is not a zero-shot comparison for the proposed method. The Ours row uses support-set labels to build the class and property caches, and the MPG property tokens are trained on the support set with the contrastive loss of Eq. (5); only the cache fine-tuning is omitted. Therefore the claim that 'direct text-image matching based on class tokens is not enough' is not established by this table. A proper control would evaluate an untrained MPG with no support-set adaptation, or the table should be relabeled as a few-shot comparison against same-shot methods.
  2. [Figs. 5-6, Table I, Table V] The paper's central claim of superiority across 11 datasets rests entirely on the plots in Figs. 5 and 6, which contain no numerical values, no error bars, and no statement of the number of seeds or variance. Given the modest reported margin on ImageNet (66.40 vs. 65.51 for Tip-Adapter-F in Table I), the reader cannot assess whether the cross-dataset gains are statistically meaningful. Additionally, Table I reports Ours=66.40 for ImageNet 16-shot while Table V reports Ours=66.45 for the same setting with ResNet50; this inconsistency should be resolved or explained.
  3. [Section V-C, Eq. (4), Table III] The property retrieval in Eq. (4) ranks LLM description clusters using cosine similarity between CLIP class tokens and text embeddings of the descriptions, which is exactly the coarse class-token representation that Section II argues suffers from 'fine-grained alignment deficiency.' The ablations in Table III vary only the score combination and never vary the retrieval source, so the gain of S_mp-cache over S_cls-cache (65.60 vs. 65.09) is also consistent with an alternative explanation: any extra trainable tokens with class-level supervision could provide a small cache benefit. To support the claim that dominant-property selection (rather than merely adding tokens or descriptions) drives the improvement, the authors should ablate the retrieval: for example, compare top-M cluster selection against random clusters, against using all LLM descriptions without pruning, and against retrieval performed with patch-level or property-token features instead of class tokens.
  4. [Section V-D, Eq. (5)] The InfoNCE loss in Eq. (5) is written in a mathematically invalid form. The numerator exp(f_mp^ij · W_p^ij^T / tau) is a scalar, while the denominator exp(f_mp^ij · [W_p^ij, W_hn^ij, W_gn^ij]^T / tau) denotes a vector because of the concatenation operation, making the logarithm of their ratio undefined. The intended expression is presumably exp(positive / tau) divided by the sum of exp(positive / tau), exp(hard negatives / tau), and exp(general negatives / tau). Since this loss is the central training objective for the property tokens, the formula needs to be corrected.
minor comments (6)
  1. [Section V-B] The text refers to 'contrastive learning objectives (Section V-B)' but the contrastive learning strategy is described in Section V-D; the cross-reference should be corrected.
  2. [Abstract] The phrase 'contrastive language-image pertaining' should be 'contrastive language-image pre-training'; the same typo appears in the Introduction.
  3. [Section II] The phrase 'dub this phenomenon asfine-grained alignment deficiency' has an obvious spacing error and should read 'as fine-grained alignment deficiency.'
  4. [Section V-C] The sentence 'Last but not least, the clustering also respects the fact that different categories share similar properties' is unclear; it is not obvious how K-means clustering over all descriptions encodes the notion that categories share properties, and this point should be clarified.
  5. [Section V-E and Table VI] The learned alpha and beta values in Table VI are reported for only four datasets, and the text says 'Larger alpha indicates more weights on property tokens' without discussing whether the sum is normalized; stating the exact score combination used at inference would improve reproducibility.
  6. [Section VI.B] The hyperparameters include several fixed choices (cluster count equal to half the number of classes, positive:negative ratio 1:100, hard:general ratio schedule, temperature 0.3, M=3) that are dataset-dependent; the sensitivity analysis only varies M (1, 3, 5) and does not test the other choices, so the robustness claim in Section VII.C.c should be narrowed accordingly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the property supervision is externally sourced (LLM text plus CLIP similarity on support images), the reported accuracies are on held-out test splits, and the only self-citations are non-load-bearing related work.

full rationale

BCT-CLIP's derivation chain does not reduce to its inputs at the equation level. The property descriptions are generated by an LLM and ranked by Eq. 4 using cosine similarity between CLIP class-token features and text-cluster embeddings; these descriptions then define positives and negatives in the InfoNCE loss Eq. 5. This makes the property-token supervision depend on CLIP's class-token alignment, but it does not make the final classification accuracy an algebraic or statistical consequence of that ranking: the MPG output f_mp is a learned function of patch features and learnable queries, trained on support images, and evaluated on held-out query images. The claim that Eq. 4 is internally inconsistent with the paper's 'fine-grained alignment deficiency' argument is a correctness and robustness concern, not a circular reduction. Hyperparameters such as M=3, the cluster count, and negative ratios are selected on a validation subset, and the reported numbers are on standard test splits, so no fitted parameter is renamed as a prediction. The two self-citations ([6], [7]) appear only in a list of related few-shot methods and carry none of the paper's load-bearing assumptions. No uniqueness theorem, ansatz, or renamed known result is imported from prior work by the same authors. The absence of a retrieval-source ablation is an interpretability gap, but it is not evidence that any predicted accuracy is equivalent by construction to the method's inputs.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The central claim rests on the assumption that LLM-generated descriptions, selected via CLIP similarity of coarse class tokens, provide trustworthy supervision for learned property tokens. The main free parameters are architecture and sampling choices tuned on ImageNet validation. No new physical entities are introduced; property tokens are learned internal representations.

free parameters (6)
  • number of property tokens M = 3
    Selected on a 5K validation set from ImageNet training data (Section VII-C-c) and applied to all datasets.
  • number of K-means clusters = half the number of categories
    Set in Section VI-B; the cluster count is dataset-dependent and hand-set.
  • positive:negative sampling ratio = 1:100
    Set in Section VI-B for contrastive learning.
  • hard:general negative ratio = starts 1:9, rises to 4:6
    Set in Section VI-B.
  • InfoNCE temperature = 0.3
    Set in Section VII-C-c.
  • MPG layers L = 2
    Set in Section VI-B.
assumptions (4)
  • domain assumption CLIP text-embedding space supports meaningful K-means clustering of property descriptions, and CLIP visual-text cosine similarity on class tokens ranks property clusters for a class.
    Used throughout Section V-C to select dominating properties; relies on the same class-token alignment the paper claims is deficient.
  • domain assumption Object properties are localized in image patches, so learnable queries with cross-attention to patch features can extract them.
    Motivates the MPG design in Section V-B.
  • domain assumption LLM-generated descriptions of class names, after clustering and selection, are faithful enough to supervise property-token learning.
    Section V-C; if descriptions are noisy or unfaithful, contrastive learning misaligns tokens.
  • standard math InfoNCE loss and cache-model scoring formulas from Tip-Adapter are valid for this setting.
    Assumed from prior work [15], [39] in Sections IV and V-D.
invented entities (1)
  • Property tokens (f_mp)
    purpose: Learned per-image vectors aligned to LLM-selected class properties, added to class token for classification.
    Internal learned representations; no falsifiable external handle beyond their role in the proposed pipeline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Class Tokens: LLM-guided Dominant Property Mining for Few-shot Classification." pith.science (2026). https://pith.science/paper/T3RCOM4I

@misc{pith2026250720511,
  author       = {Pith},
  title        = {Pith review of: Beyond Class Tokens: LLM-guided Dominant Property Mining for Few-shot Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T3RCOM4I}},
  note         = {Machine review of arXiv:2507.20511}
}
read the original abstract

Few-shot Learning (FSL), which endeavors to develop the generalization ability for recognizing novel classes using only a few images, faces significant challenges due to data scarcity. Recent CLIP-like methods based on contrastive language-image pertaining mitigate the issue by leveraging textual representation of the class name for unseen image discovery. Despite the achieved success, simply aligning visual representations to class name embeddings would compromise the visual diversity for novel class discrimination. To this end, we proposed a novel Few-Shot Learning (FSL) method (BCT-CLIP) that explores \textbf{dominating properties} via contrastive learning beyond simply using class tokens. Through leveraging LLM-based prior knowledge, our method pushes forward FSL with comprehensive structural image representations, including both global category representation and the patch-aware property embeddings. In particular, we presented a novel multi-property generator (MPG) with patch-aware cross-attentions to generate multiple visual property tokens, a Large-Language Model (LLM)-assistant retrieval procedure with clustering-based pruning to obtain dominating property descriptions, and a new contrastive learning strategy for property-token learning. The superior performances on the 11 widely used datasets demonstrate that our investigation of dominating properties advances discriminative class-specific representation learning and few-shot classification.

Figures

Figures reproduced from arXiv: 2507.20511 by the authors.

Figure 1
Figure 1. Comparison with existing CLIP-based FSL methods. Instead of using one single prototype for novel class, our approach introduces a multi-property framework, significantly enriching the representation of novel classes. This is enabled by our efficient Multi-Property Generator (MPG) and contrastive learning, which leverages properties retrieved from Large Language Models (LLMs). this, a generalized powerful pretrained … view at source ↗
Figure 2
Figure 2. The framework of BCT-CLIP. We introduce a property retrieval process based on LLMs. During training, we employ contrastive learning to generate property tokens, which are aligned with the selected property descriptions. Additionally, we construct caches based on the prototypes that are derived from the training set to enable direct classification. V. METHOD A. Problem Definition and Overview Few-shot learning focuse… view at source ↗
Figure 3
Figure 3. The Structure of Multi-property Generator. It consists of a cross￾attention layer, a group-wise Feed Forward Network (FFN), and two norm layers, with residual connections added in between. The input part is made up of patch tokens of the image as K, V, and randomly initialized learnable tokens as Q. of information, thereby facilitating more effective clustering. For content generation, we provide class names and req… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The process of selecting positive and general/hard negative properties from clusters. For each property token, its ground truth positive and negative descriptions are randomly sampled from the corresponding clusters in each training iteration. Here, green, blue and red…
Figure 5
Figure 5. Figure 5: Comparison(%) of different SOTA methods under various few-shot settings on 10 downstream tasks. Our method outperforms previous methods on [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Average results on 11 downstream tasks. Our method consistently [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Comparison between our selected descriptions and those generated [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

58 extracted references · 33 canonical work pages

  1. [1]

    Siamese neural networks for one-shot image recognition,

    G. Koch, R. Zemel, R. Salakhutdinovet al., “Siamese neural networks for one-shot image recognition,” inICML deep learning workshop, vol. 2, no. 1. Lille, 2015, pp. 1–30

    work page 2015

  2. [2]

    Matching networks for one shot learning,

    O. Vinyals, C. Blundell, T. Lillicrap, D. Wierstraet al., “Matching networks for one shot learning,” pp. 3630–3638, 2016

    work page 2016

  3. [3]

    Prototypical networks for few-shot learning,

    J. Snell, K. Swersky, and R. Zemel, “Prototypical networks for few-shot learning,”Advances in neural information processing systems, vol. 30, 2017

    2017

  4. [4]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III

    2015

  5. [5]

    Meta r- cnn: Towards general solver for instance-level low-shot learning,

    X. Yan, Z. Chen, A. Xu, X. Wang, X. Liang, and L. Lin, “Meta r- cnn: Towards general solver for instance-level low-shot learning,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 9577–9586

    work page 2019

  6. [6]

    Few-shot object detection with attention-rpn and multi-relation detector,

    Q. Fan, W. Zhuo, C.-K. Tang, and Y .-W. Tai, “Few-shot object detection with attention-rpn and multi-relation detector,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 4013–4022

    work page 2020

  7. [7]

    Mining latent classes for few-shot segmentation,

    L. Yang, W. Zhuo, L. Qi, Y . Shi, and Y . Gao, “Mining latent classes for few-shot segmentation,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 8721–8730

    work page 2021

  8. [8]

    Imagenet classification with deep convolutional neural networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional neural networks,”Advances in neural informa- tion processing systems, vol. 25, 2012

    2012

Show all 58 references
  1. [9]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,”arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  2. [10]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778

    2016

  3. [11]

    Learning to prompt for vision- language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to prompt for vision- language models,”International Journal of Computer Vision, vol. 130, no. 9, pp. 2337–2348, 2022

    2022

  4. [12]

    Conditional prompt learning for vision-language models,

    K. Zhou, J. Yang, C. Loy, and Z. Liu, “Conditional prompt learning for vision-language models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 16 816–16 825

    2022

  5. [13]

    What does a platypus look like? generating customized prompts for zero-shot image classification,

    S. Pratt, I. Covert, R. Liu, and A. Farhadi, “What does a platypus look like? generating customized prompts for zero-shot image classification,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 15 691–15 701

    2023

  6. [14]

    Chatgpt-powered hierarchical comparisons for image classification,

    Z. Ren, Y . Su, and X. Liu, “Chatgpt-powered hierarchical comparisons for image classification,”Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  7. [15]

    Tip-adapter: Training-free clip-adapter for better vision-language modeling,

    R. Zhang, R. Fang, W. Zhang, P. Gao, K. Li, J. Dai, Y . Qiao, and H. Li, “Tip-adapter: Training-free clip-adapter for better vision-language modeling,”arXiv preprint arXiv:2111.03930, 2021

    2021 arXiv

  8. [16]

    Sus-x: Training-free name- only transfer of vision-language models,

    V . Udandarao, A. Gupta, and S. Albanie, “Sus-x: Training-free name- only transfer of vision-language models,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 2725–2736

    2023

  9. [17]

    Partial-tuning based mixed-modal prototypes for few-shot classification,

    Y . Su, X. Liu, Y . Zhao, R. Hong, and M. Wang, “Partial-tuning based mixed-modal prototypes for few-shot classification,”IEEE Transactions on Multimedia, vol. 26, pp. 9175–9186, 2024

    2024

  10. [18]

    Springer, 2015, pp. 234–241

    2015

  11. [19]

    Meta networks,

    T. Munkhdalai and H. Yu, “Meta networks,” inProceedings of the 34th International Conference on Machine Learning-Volume 70. JMLR. org, 2017, pp. 2554–2563

    2017

  12. [20]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 10 684–10 695

    2022

  13. [21]

    On first-order meta-learning algorithms,

    A. Nichol, J. Achiam, and J. Schulman, “On first-order meta-learning algorithms,”arXiv preprint arXiv:1803.02999, 2018

    2018 arXiv

  14. [22]

    Attribute-guided feature learning for few- shot image recognition,

    Y . Zhu, W. Min, and S. Jiang, “Attribute-guided feature learning for few- shot image recognition,”IEEE Transactions on Multimedia, vol. 23, pp. 1200–1209, 2020

    2020

  15. [23]

    Revisiting fine-tuning for few-shot learn- ing,

    A. Nakamura and T. Harada, “Revisiting fine-tuning for few-shot learn- ing,”arXiv preprint arXiv:1910.00216, 2019

    1910 arXiv

  16. [24]

    A closer look at few-shot classification,

    W.-Y . Chen, Y .-C. Liu, Z. Kira, Y .-C. Wang, and J.-B. Huang, “A closer look at few-shot classification,” inInternational Conference on Learning Representations, 2019

    2019

  17. [25]

    Meta-baseline: Exploring simple meta-learning for few-shot learning,

    Y . Chen, Z. Liu, H. Xu, T. Darrell, and X. Wang, “Meta-baseline: Exploring simple meta-learning for few-shot learning,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 9062–9071

    2021

  18. [26]

    A baseline for few-shot image classification,

    G. S. Dhillon, P. Chaudhari, A. Ravichandran, and S. Soatto, “A baseline for few-shot image classification,” 2019

    2019

  19. [27]

    Self-training for few-shot transfer across extreme task differences,

    C. P. Phoo and B. Hariharan, “Self-training for few-shot transfer across extreme task differences,”arXiv:2010.07734, 2020

    2010 arXiv

  20. [28]

    Universal representation learning from multiple domains for few-shot classification,

    W.-H. Li, X. Liu, and H. Bilen, “Universal representation learning from multiple domains for few-shot classification,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 9526– 9535

    2021

  21. [29]

    Visual-language prompt tuning with knowledge-guided context optimization,

    H. Yao, R. Zhang, and C. Xu, “Visual-language prompt tuning with knowledge-guided context optimization,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 6757–6767

    2023

  22. [30]

    Prompt-aligned gradient for prompt tuning,

    B. Zhu, Y . Niu, Y . Han, Y . Wu, and H. Zhang, “Prompt-aligned gradient for prompt tuning,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 15 659–15 669

    2023

  23. [31]

    Visual classification via description from large language models,

    S. Menon and C. V ondrick, “Visual classification via description from large language models,” 2023

    2023

  24. [32]

    Clip-adapter: Better vision-language models with feature adapters,

    P. Gao, S. Geng, R. Zhang, T. Ma, R. Fang, Y . Zhang, H. Li, and Y . Qiao, “Clip-adapter: Better vision-language models with feature adapters,” International Journal of Computer Vision, vol. 132, no. 2, pp. 581–595, 2024

    2024

  25. [33]

    Infinite mixture prototypes for few-shot learning,

    K. Allen, E. Shelhamer, H. Shin, and J. Tenenbaum, “Infinite mixture prototypes for few-shot learning,” inInternational conference on ma- chine learning. PMLR, 2019, pp. 232–241

    2019

  26. [34]

    Multi- prototype few-shot learning in histopathology,

    J. Deuschel, D. Firmbach, C. I. Geppert, M. Eckstein, A. Hartmann, V . Bruns, P. Kuritcyn, J. Dexl, D. Hartmann, D. Perrinet al., “Multi- prototype few-shot learning in histopathology,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 620– 628

    2021

  27. [35]

    Local descriptor-based multi-prototype network for few-shot learning,

    H. Huang, Z. Wu, W. Li, J. Huo, and Y . Gao, “Local descriptor-based multi-prototype network for few-shot learning,”Pattern Recognition, vol. 116, p. 107935, 2021

    2021

  28. [36]

    Token contrast for weakly- supervised semantic segmentation,

    L. Ru, H. Zheng, Y . Zhan, and B. Du, “Token contrast for weakly- supervised semantic segmentation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 3093–3102

    2023

  29. [37]

    Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, XXXX 2025 11 models,

    J. Li, D. Li, S. Savarese, and S. Hoi, “Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, XXXX 2025 11 models,” inInternational conference on machine learning. PMLR, 2023, pp. 19 730–19 742

    2025

  30. [38]

    Flamingo: a visual language model for few-shot learning,

    J.-B. Alayrac, J. Donahue, P. Luc, A. Miech, I. Barr, Y . Hasson, K. Lenc, A. Mensch, K. Millican, M. Reynoldset al., “Flamingo: a visual language model for few-shot learning,”Advances in neural information processing systems, vol. 35, pp. 23 716–23 736, 2022

    2022

  31. [39]

    Ofa: Unifying architectures, tasks, and modalities through a simple sequence-to-sequence learning framework,

    P. Wang, A. Yang, R. Men, J. Lin, S. Bai, Z. Li, J. Ma, C. Zhou, J. Zhou, and H. Yang, “Ofa: Unifying architectures, tasks, and modalities through a simple sequence-to-sequence learning framework,” inInternational conference on machine learning. PMLR, 2022, pp. 23 318–23 340

    2022

  32. [40]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clarket al., “Learning transferable visual models from natural language supervision,” inInternational conference on machine learning. PMLR, 2021, pp. 8748–8763

    2021

  33. [41]

    Language mod- els are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askellet al., “Language mod- els are few-shot learners,”Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020

    1901

  34. [42]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in2009 IEEE Conference on Computer Vision and Pattern Recognition, 2009, pp. 248–255

    2009

  35. [43]

    3d object representations for fine-grained categorization,

    J. Krause, M. Stark, J. Deng, and L. Fei-Fei, “3d object representations for fine-grained categorization,” in2013 IEEE International Conference on Computer Vision Workshops, 2013, pp. 554–561

    2013

  36. [44]

    Ucf101: A dataset of 101 human actions classes from videos in the wild,

    K. Soomro, A. R. Zamir, and M. Shah, “Ucf101: A dataset of 101 human actions classes from videos in the wild,” 2012. [Online]. Available: https://arxiv.org/abs/1212.0402

    2012 arXiv

  37. [45]

    Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories,

    L. Fei-Fei, R. Fergus, and P. Perona, “Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories,” in2004 Conference on Computer Vision and Pattern Recognition Workshop, 2004, pp. 178–178

    2004

  38. [46]

    Automated flower classification over a large number of classes,

    M.-E. Nilsback and A. Zisserman, “Automated flower classification over a large number of classes,” in2008 Sixth Indian conference on computer vision, graphics & image processing. IEEE, 2008, pp. 722–729

    2008

  39. [47]

    Sun database: Large-scale scene recognition from abbey to zoo,

    J. Xiao, J. Hays, K. A. Ehinger, A. Oliva, and A. Torralba, “Sun database: Large-scale scene recognition from abbey to zoo,” in2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010, pp. 3485–3492

    2010

  40. [48]

    Describing textures in the wild,

    M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, and A. Vedaldi, “Describing textures in the wild,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2014, pp. 3606–3613

    2014

  41. [49]

    Eurosat: A novel dataset and deep learning benchmark for land use and land cover classi- fication,

    P. Helber, B. Bischke, A. Dengel, and D. Borth, “Eurosat: A novel dataset and deep learning benchmark for land use and land cover classi- fication,”IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 12, no. 7, pp. 2217–2226, 2019

    2019

  42. [50]

    Fine- grained visual classification of aircraft,

    S. Maji, E. Rahtu, J. Kannala, M. Blaschko, and A. Vedaldi, “Fine- grained visual classification of aircraft,” 2013. [Online]. Available: https://arxiv.org/abs/1306.5151

    2013 arXiv

  43. [51]

    Cats and dogs,

    O. M. Parkhi, A. Vedaldi, A. Zisserman, and C. V . Jawahar, “Cats and dogs,” in2012 IEEE Conference on Computer Vision and Pattern Recognition, 2012, pp. 3498–3505

    2012

  44. [52]

    Food-101 – mining discriminative components with random forests,

    L. Bossard, M. Guillaumin, and L. Van Gool, “Food-101 – mining discriminative components with random forests,” inComputer Vision – ECCV 2014, D. Fleet, T. Pajdla, B. Schiele, and T. Tuytelaars, Eds. Cham: Springer International Publishing, 2014, pp. 446–461

    2014

  45. [53]

    Plot: Prompt learning with optimal transport for vision-language models,

    G. Chen, W. Yao, X. Song, X. Li, Y . Rao, and K. Zhang, “Plot: Prompt learning with optimal transport for vision-language models,” inICLR, 2023

    2023

  46. [54]

    Lp++: A surprisingly strong linear probe for few-shot clip,

    Y . Huang, F. Shakeri, J. Dolz, M. Boudiaf, H. Bahig, and I. Ben Ayed, “Lp++: A surprisingly strong linear probe for few-shot clip,” inPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 23 773–23 782

    2024

  47. [55]

    A closer look at the few-shot adaptation of large vision-language models,

    J. Silva-Rodriguez, S. Hajimiri, I. Ben Ayed, and J. Dolz, “A closer look at the few-shot adaptation of large vision-language models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 23 681–23 690

    2024

  48. [56]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  49. [57]

    Do imagenet clas- sifiers generalize to imagenet?

    B. Recht, R. Roelofs, L. Schmidt, and V . Shankar, “Do imagenet clas- sifiers generalize to imagenet?” inInternational conference on machine learning. PMLR, 2019, pp. 5389–5400

    2019

  50. [58]

    Learning robust global representations by penalizing local predictive power,

    H. Wang, S. Ge, Z. Lipton, and E. P. Xing, “Learning robust global representations by penalizing local predictive power,”Advances in neural information processing systems, vol. 32, 2019

    2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.