Pith. sign in

REVIEW 4 major objections 5 minor 67 references

Beyond Graph Model: Reliable VLM Fine-Tuning via Random Graph Adapter

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read VRGAdapter claims that modeling each class as a Gaussian distribution over LLM-generated descriptions and propagating those distributions through a class graph yields a more general and more accurate VLM adapter than deterministic graph…

desk verdict The headline SOTA numbers are mostly supplied by the MoCo/DINO ensemble, not the graph adapter; the controlled gain over GraphAdapter is real but modest and under-reported. read the letter →

arxiv 2507.10355 v1 pith:3CSG6S35 submitted 2025-07-14 cs.CV

Bo Jiang , Xueyang Ze , Beibei Wang , Xixi Wang , Xixi Wan , Bin Luo This is my paper · ORCID
classification cs.CV
keywords vision-languagemodelsfew-shotclassificationtextualadapterrandomgraphGaussiandistributionneuralnetworkuncertainty-guidedfusionout-of-distributiongeneralization
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

Diverse text descriptions of one category are not noise; they are signal, and the paper's aim is to put that signal into the adapter rather than average it away. To do this, VRGAdapter represents each class as a Gaussian distribution estimated from many LLM-generated descriptions, builds a graph whose edges connect class means, and propagates both means and variances across the graph before sampling refined text prototypes. The paper claims this random-graph adapter is strictly more general than the deterministic GraphAdapter, recovering it when each class has a single description, and that it improves few-shot classification consistently across 11 benchmarks, e.g., 71.35% on ImageNet-1K 16-shot with a ResNet-50 backbone. It also adds an uncertainty-guided fusion that weights predictions from CLIP and two auxiliary self-supervised encoders by per-sample confidence, which the ablations show contributes on top of the graph adapter. If right, the work gives a way to fine-tune frozen VLMs that captures semantic variation and inter-class structure together, without full fine-tuning or extra synthetic training data.

What carries the argument

The load-bearing object is the Vertex Random Knowledge Graph (VRKG), a class-level graph whose nodes are Gaussian distributions rather than deterministic vectors. Its edges are fixed in advance as cosine similarities between class-mean text features, and message passing is carried out by a Gaussian-based graph convolution that propagates both the mean $\mu_i$ and the diagonal variance $\sigma_i$ of each node. The reparameterized sampling step at the end makes the whole adapter trainable by gradient descent, and the residual blend with the original mean keeps the pretrained semantic knowledge from being overwritten. This object is what lets the adapter capture two things at once: the intra-class diversity encoded in the variance, and the inter-class relationships encoded in the edges and the aggregation.

What would settle it

Train the same VRGAdapter pipeline on ImageNet-1K 16-shot but replace the edge weights with random values, with distribution-aware distances, or with learned edges; if random edges keep accuracy near 71.35%, the graph is not carrying the claimed signal, and if distribution-aware edges change accuracy substantially, the mean-cosine edge choice is the active ingredient.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that a VLM textual adapter can carry a class's semantic diversity through training by making each class node a Gaussian distribution instead of a point. The vertex random knowledge graph is built from $M=50$ LLM-generated descriptions per class: the mean $\mu_i$ and diagonal variance $\sigma_i$ of node $i$ are estimated from the encoded descriptions, and the edge weight between nodes $i$ and $j$ is fixed as $A_{ij} = \cos(\mu_i,\mu_j)$. Two layers of Gaussian-based graph convolution then update both parameters of every node, after which a reparameterized sample $z_i = \mu_i^{(L)} + \epsilon \odot \sqrt{\sigma_i^{(L)}}$ produces the adapted textual prototype, blended with the original mean by the residual weight $\alpha$. Complementing the graph adapter, the Uncertainty-guided Multi-branch Fusion (UMF) scheme weights each branch's logits by the normalized kurtosis of its prediction distribution, so a model that is more confident on a given sample contributes more to the fused prediction. The paper reports that this pipeline outperforms prior adapters on all 11 few-shot benchmarks and on ImageNet-V2 and ImageNet-Sketch, and states that GraphAdapter is the special case with $M=1$, which is the basis for calling VRGAdapter the more general adapter solution.

Load-bearing premise

The graph-propagation step assumes that how similar two classes' average descriptions are tells you how much those classes should share information during message passing; if that similarity is misleading, the adapter will blend class knowledge using the wrong weights.

Editorial extensions

If this is right

  • Using more LLM descriptions per class improves accuracy up to a saturation point around $M=20$, so the variance signal, not just the added compute, is what helps.
  • Because GraphAdapter is the $M=1$ limit of VRGAdapter, any deterministic graph adapter can be upgraded to the distributional version by generating multiple descriptions and reusing the same propagation code.
  • The UMF ablation shows that confidence-weighted fusion beats both no fusion and simple averaging of the same CLIP and auxiliary branches, indicating that per-sample uncertainty is a usable signal for ensembling frozen encoders.
  • The paper reports larger gains at higher shot counts and improved transfer to ImageNet-V2 and ImageNet-Sketch, so the method's benefits are not limited to the training distribution.

Reading between the lines

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

  • A natural next step the paper does not run: replace the deterministic edge weights with a similarity between the full Gaussian nodes, such as a Bhattacharyya or Wasserstein kernel, and check whether the gains grow; this would separate the contribution of the graph structure from the contribution of the mean features that define the edges.
  • Because the kurtosis weighting is defined on any logit vector, the same UMF rule could fuse more than two auxiliary models, or be tested as a general ensembling method for frozen encoders on other downstream tasks such as open-vocabulary detection or retrieval.
  • The method's performance should be sensitive to the quality and diversity of the LLM-generated descriptions; a controlled experiment that downsamples the descriptions or replaces the LLM with simple template paraphrases would reveal how much of the gain is due to the distributional model rather than the particular prompts.
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 / 5 minor

Summary. The paper proposes VRGAdapter, a textual adapter for fine-tuning vision-language models in the few-shot setting. Each class is represented as a Gaussian distribution estimated from M LLM-generated textual descriptions, and a Vertex Random Knowledge Graph is used to propagate both means and variances through a Gaussian-based graph convolution. A reparameterized sampling step converts the refined distributions into class prototypes, and an Uncertainty-guided Multi-branch Fusion (UMF) scheme combines CLIP predictions with MoCo and DINO auxiliary branches using a kurtosis-based confidence weight. The authors report state-of-the-art few-shot accuracies on ImageNet-1K, ten other downstream datasets, and two out-of-distribution benchmarks, and they claim that VRGAdapter encompasses traditional graph adapters as a special case.

Significance. If the central claims are substantiated, the paper would contribute a useful direction: representing class-level textual features as distributions and propagating them on a graph is a natural extension of deterministic adapters, and the UMF scheme is a practical way to combine complementary encoders. The paper includes a broad experimental sweep across 11 benchmarks, a hyperparameter sensitivity analysis, a controlled comparison in Table V, and t-SNE visualizations of the learned distributions. However, the current evidence does not cleanly attribute the headline margins to the random-graph mechanism: the main comparison is confounded by the auxiliary branches, the controlled gains are small and reported without error bars, and the probabilistic propagation step has a technical issue with signed edge weights. The work is promising but needs revision before the stated claims are reliable.

major comments (4)
  1. [V-B (Table I) and V-C (Table V)] The headline comparison is confounded. In Table I the VRGAdapter row is the full system with MoCo and DINO auxiliary branches and the UMF fusion described in Section IV, while the GraphAdapter row is a single-branch CLIP adapter. The reported gains of 2.38-5.65 points over GraphAdapter are therefore not a clean measure of the proposed graph module. The controlled comparison in Table V, in which the common base is CLIP+MoCo+DINO, shows a VRGAdapter-versus-GraphAdapter gain of only 0.52-1.76 points on ImageNet-1K and 1.13-3.84 points on DTD. The paper should present the controlled margins as the primary evidence for VRGAdapter and clearly separate the ensemble-level results from the adapter-level results.
  2. [III-B, Eqs. (7)-(8)] The probabilistic message passing is not a valid covariance update with signed edges. Equation (6) defines A_ij as a cosine similarity that can be negative, and Eq. (8) multiplies neighbor variances by these signed coefficients before applying ReLU. A negative coefficient can make the updated diagonal variance non-positive before the nonlinearity, and the update is not a convex combination of variances, so the propagated sigma_i^(L) cannot be interpreted as the covariance of a Gaussian. This matters because the paper's central claim is that VRGAdapter performs probabilistic graph propagation. The authors should either restrict the graph to non-negative affinities, such as a cosine similarity clipped at zero or a positive kernel, or justify why the signed update still yields a meaningful distribution; an ablation over edge constructions would directly address this concern.
  3. [III-A and Introduction] The claim that GraphAdapter is a special case for M=1 is not supported by the described models. GraphAdapter [2] builds a dual knowledge graph with both textual and visual class nodes, whereas VRGAdapter with M=1 is a text-only class graph whose nodes are degenerate Gaussians; the two architectures do not coincide. In addition, the alpha in Eq. (11) and the reparameterized sampling in Eq. (10) mean that even at M=1 the output is not identical to GraphAdapter's update. The paper should either provide a formal reduction or weaken the statement to 'a deterministic single-prototype text-only GCN is a limiting case of VRGAdapter.'
  4. [V-C (Table V)] The controlled comparison in Table V is reported as single point estimates without random seeds or error bars. Because the ImageNet gains over GraphAdapter are as small as 0.52 points at the 16-shot setting, it is not possible to determine whether the random-graph component yields a statistically meaningful improvement. Please report means and standard deviations over at least three seeds and release the code so that the experiments can be reproduced.
minor comments (5)
  1. [III-B, Eqs. (7)-(8)] The activation function is denoted sigma(·), which is the same symbol used for the variance parameters sigma_i; please rename one of the two to avoid confusion.
  2. [III-C] The inference procedure is unspecified: Eq. (10) describes sampling for the reparameterization trick, but the paper does not state whether test-time prediction uses one sample, the posterior mean, or multiple samples.
  3. [V-C, Table IV] The ablation notation is unclear: the row labeled 'AUX' is described in the text as averaging auxiliary predictions, while the row 'AUX+UMF' uses the kurtosis weighting; please define both rows explicitly.
  4. [Tables I, III, and VI] There are small numerical inconsistencies between tables: GraphAdapter source accuracy is 65.70 in Table I but 64.94 in Table III for ResNet-50, and 73.68 in Table VI but 73.40 in Table III for ViT-B/16; please reconcile these values or explain the differences.
  5. [IV-B, Eq. (17)] The kurtosis measure is applied to a C-dimensional logit vector with an exponent lambda, but no justification is given for why this quantity is a reliable confidence measure for small C, such as EuroSAT with 10 classes; please add a short discussion or an ablation on the number of classes.

Circularity Check

0 steps flagged · score 2.0 of 10

No substantive circularity in VRGAdapter; its equations, ablations, and external benchmarks ground the central claim, with only a minor non-load-bearing self-citation to HeGraphAdapter [27].

full rationale

The claimed gains are not forced by construction. The VRGAdapter pipeline defines class distributions from LLM-generated descriptions (Eqs. 4-5), propagates them with a Gaussian GCN (Eqs. 7-8), samples reparameterized text features (Eq. 10), and blends the result with the original mean (Eq. 11); each step is an independently defined operation whose output is tested on held-out benchmarks. The statement that GraphAdapter is the M=1 special case is a mathematical reduction, not a self-referential derivation of the accuracy numbers. The UMF kurtosis weight is computed from the same logits being fused, but this is a heuristic confidence weighting rather than a fitted parameter renamed as a prediction, and the fused pipeline is evaluated externally. Tables I and II compare the full auxiliary-equipped system with baselines that lack auxiliary branches, and Table V shows smaller controlled margins; that is an attribution and experimental-design concern, not circularity. The only overlapping-author citation is [27] (HeGraphAdapter), used as related work and not as a load-bearing justification, so the low score reflects the minor self-citation rather than any circular derivation.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The ledger shows that the method rests on several domain-specific assumptions about Gaussianity, graph edge construction, and kurtosis as confidence, none of which are verified beyond task performance. The free hyperparameters are tuned on validation sets, which is standard but adds five hand-chosen constants.

free parameters (5)
  • alpha (α) = 0.7
    Trade-off between original and adapter-enhanced representations in Eq. 11; tuned on validation set (Section V.A.2).
  • lambda (λ) = 0.4
    Exponent in kurtosis confidence metric in Eq. 17; tuned on validation set.
  • beta (β) = 0.5
    Fusion weight between CLIP and auxiliary branches in Eq. 19; tuned on validation set.
  • M = 50
    Number of LLM-generated descriptions per class; chosen via hyperparameter search, performance saturates around M=20 (Fig. 4).
  • hidden_dim = 16
    Hidden feature dimension of the two-layer Gaussian GCN; no ablation reported.
assumptions (6)
  • domain assumption CLIP text embeddings of diverse descriptions for a class are adequately summarized by a Gaussian distribution (mean and diagonal covariance) in Eqs. 3-5.
    No goodness-of-fit test or justification that the embeddings are unimodal and Gaussian.
  • domain assumption Edge weights A_ij = cos(μ_i, μ_j) in Eq. 6 capture inter-class relationships well enough for message passing.
    Uses only means, ignoring the variances; no sensitivity analysis.
  • domain assumption The Gaussian graph convolution operations in Eqs. 7-8, adopted from [44], preserve the Gaussian distributional form and provide useful representations.
    Assumes the prior derivation of the Gaussian GCN is applicable to VLM text features.
  • standard math The reparameterization trick in Eq. 10 yields valid gradients for training.
    Standard result from [45].
  • ad hoc to paper Kurtosis of the class logit vector in Eq. 17 is a valid and reliable measure of prediction confidence.
    No theoretical or empirical justification beyond an ablation; other confidence measures are not compared.
  • domain assumption LLM-generated descriptions (Eq. 1) provide a diverse and representative sample of the semantic variation of each class.
    Depends on the unspecified LLM and prompt template; no analysis of the diversity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Graph Model: Reliable VLM Fine-Tuning via Random Graph Adapter." pith.science (2026). https://pith.science/paper/3CSG6S35

@misc{pith2026250710355,
  author       = {Pith},
  title        = {Pith review of: Beyond Graph Model: Reliable VLM Fine-Tuning via Random Graph Adapter},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3CSG6S35}},
  note         = {Machine review of arXiv:2507.10355}
}
read the original abstract

Textual adapter-based tuning methods have shown significant potential in transferring knowledge from pre-trained Vision-Language Models (VLMs) to downstream tasks. Existing works generally employ the deterministic textual feature adapter to refine each category textual representation. However, due to inherent factors such as different attributes and contexts, there exists significant diversity in textual descriptions for each category. Such description diversity offers rich discriminative semantic knowledge that can benefit downstream visual learning tasks. Obviously, traditional deterministic adapter model cannot adequately capture this varied semantic information. Also, it is desirable to exploit the inter-class relationships in VLM adapter. To address these issues, we propose to exploit random graph model into VLM adapter and develop a novel Vertex Random Graph Adapter (VRGAdapter). VRGAdapter first models the inherent diverse descriptions of each category and inter-class relationships of different categories simultaneously by leveraging a Vertex Random Knowledge Graph (VRKG) model. Then, it employs probabilistic message propagation on VRKG to learn context-aware distribution representation for each class node. Finally, it adopts a reparameterized sampling function to achieve textual adapter learning. Note that, VRGAdapter provides a more general adapter solution that encompasses traditional graph-based adapter as a special case. In addition, to enable more robust performance for downstream tasks, we also introduce a new Uncertainty-guided Multi-branch Fusion (UMF) scheme that dynamically integrates multiple pre-trained models for ensemble prediction. Extensive experiments on multiple benchmark datasets demonstrate the effectiveness of our approach.

Figures

Figures reproduced from arXiv: 2507.10355 by the authors.

Figure 1
Figure 1. Comparison of different textual branch adaptation strategies for [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of our proposed approach. The upper part illustrates the Vertex Random Graph Adapter (VRGAdapter) process: diverse descriptions are [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Performance comparison (%) of different SOTA methods on few-shot classification, i.e., 1-/2-/4-/8-/16-shot, across 11 downstream tasks. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Ablation studies (%) for hyperparameters on ImageNet-1K. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: 2D t-SNE visualization of class distributions before and after applying VRGAdapter. Original class centers (circles) and optimized class centers [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

67 extracted references · 53 canonical work pages

  1. [44]

    Robust graph convolutional networks against adversarial attacks,

    D. Zhu, Z. Zhang, P. Cui, and W. Zhu, “Robust graph convolutional networks against adversarial attacks,” in Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019, pp. 1399–1407

    work page 2019

  2. [2]

    Graphadapter: Tuning vision-language models with dual knowledge graph,

    X. Li, D. Lian, Z. Lu, J. Bai, Z. Chen, and X. Wang, “Graphadapter: Tuning vision-language models with dual knowledge graph,” Advances in Neural Information Processing Systems , vol. 36, 2023

    work page 2023

  3. [1]

    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,” in ICCV, 2023, pp. 15 691–15 701

    work page 2023

  4. [3]

    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. Clark et al., “Learning transferable visual models from natural language supervision,” in ICML, 2021, pp. 8748–8763

    2021

  5. [4]

    Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation,

    J. Li, D. Li, C. Xiong, and S. Hoi, “Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation,” in International conference on machine learning . PMLR, 2022, pp. 12 888–12 900

    2022

  6. [5]

    Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,

    J. Li, D. Li, S. Savarese, and S. Hoi, “Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,” in International conference on machine learning . PMLR, 2023, pp. 19 730–19 742

    2023

  7. [6]

    Calip: Zero-shot enhancement of clip with parameter-free attention,

    Z. Guo, R. Zhang, L. Qiu, X. Ma, X. Miao, X. He, and B. Cui, “Calip: Zero-shot enhancement of clip with parameter-free attention,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 1, 2023, pp. 746–754

    work page 2023

  8. [7]

    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,” inICCV, 2023, pp. 2725–2736

    work page 2023

Show all 67 references
  1. [8]

    Learning to prompt for vision- language models,

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

    2022

  2. [9]

    How does fine-tuning impact out-of-distribution de- tection for vision-language models?

    Y . Ming and Y . Li, “How does fine-tuning impact out-of-distribution de- tection for vision-language models?” International Journal of Computer Vision, vol. 132, no. 2, pp. 596–609, 2024

    2024

  3. [10]

    Ifseg: Image-free se- mantic segmentation via vision-language model,

    S. Yun, S. H. Park, P. H. Seo, and J. Shin, “Ifseg: Image-free se- mantic segmentation via vision-language model,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 2967–2977

    2023

  4. [11]

    Robust fine-tuning of zero-shot models,

    M. Wortsman, G. Ilharco, J. W. Kim, M. Li, S. Kornblith, R. Roelofs, R. G. Lopes, H. Hajishirzi, A. Farhadi, H. Namkoong et al. , “Robust fine-tuning of zero-shot models,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 7959– 7971

    2022

  5. [12]

    Fd-align: feature discrimination alignment for fine-tuning pre-trained models in few-shot learning,

    K. Song, H. Ma, B. Zou, H. Zhang, and W. Huang, “Fd-align: feature discrimination alignment for fine-tuning pre-trained models in few-shot learning,” Advances in Neural Information Processing Systems , vol. 36, 2023

    2023

  6. [13]

    Conditional prompt learning for vision-language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Conditional prompt learning for vision-language models,” in CVPR, 2022, pp. 16 816–16 825

    2022

  7. [14]

    Prompt distribution learning,

    Y . Lu, J. Liu, Y . Zhang, Y . Liu, and X. Tian, “Prompt distribution learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 5206–5215

    2022

  8. [15]

    Bi-modality individual- aware prompt tuning for visual-language model,

    H. Yao, R. Zhang, H. Lyu, Y . Zhang, and C. Xu, “Bi-modality individual- aware prompt tuning for visual-language model,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2025

    2025

  9. [16]

    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,” IJCV, pp. 1–15, 2023

    2023

  10. [17]

    Tip-Adapter: Training-free adaption of clip for few-shot classification,

    R. Zhang, W. Zhang, R. Fang, P. Gao, K. Li, J. Dai, Y . Qiao, and H. Li, “Tip-Adapter: Training-free adaption of clip for few-shot classification,” in ECCV. Springer, 2022, pp. 493–510

    2022

  11. [18]

    Task residual for tuning vision-language models,

    T. Yu, Z. Lu, X. Jin, Z. Chen, and X. Wang, “Task residual for tuning vision-language models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 10 899–10 909

    2023

  12. [19]

    Tun- ing vision-language models with multiple prototypes clustering,

    M.-H. Guo, Y . Zhang, T.-J. Mu, S. X. Huang, and S.-M. Hu, “Tun- ing vision-language models with multiple prototypes clustering,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024

    2024

  13. [20]

    Mma: Multi-modal adapter for vision-language models,

    L. Yang, R.-Y . Zhang, Y . Wang, and X. Xie, “Mma: Multi-modal adapter for vision-language models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 826–23 837

    2024

  14. [21]

    Maple: Multi-modal prompt learning,

    M. U. Khattak, H. Rasheed, M. Maaz, S. Khan, and F. S. Khan, “Maple: Multi-modal prompt learning,” in CVPR, 2023, pp. 19 113–19 122

    2023

  15. [22]

    Bayesian prompt learn- ing for image-language model generalization,

    M. M. Derakhshani, E. Sanchez, A. Bulat, V . G. T. da Costa, C. G. Snoek, G. Tzimiropoulos, and B. Martinez, “Bayesian prompt learn- ing for image-language model generalization,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 15 237–15 246

    2023

  16. [23]

    Promp- tkd: Unsupervised prompt distillation for vision-language models,

    Z. Li, X. Li, X. Fu, X. Zhang, W. Wang, S. Chen, and J. Yang, “Promp- tkd: Unsupervised prompt distillation for vision-language models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 26 617–26 626

    2024

  17. [24]

    Prompt, generate, then cache: Cascade of foundation models makes strong few-shot learners,

    R. Zhang, X. Hu, B. Li, S. Huang, H. Deng, Y . Qiao, P. Gao, and H. Li, “Prompt, generate, then cache: Cascade of foundation models makes strong few-shot learners,” in CVPR, 2023, pp. 15 211–15 222

    2023

  18. [25]

    Textrefiner: Internal visual feature as efficient refiner for vision-language models prompt tun- ing,

    J. Xie, Y . Zhang, J. Peng, Z. Huang, and L. Cao, “Textrefiner: Internal visual feature as efficient refiner for vision-language models prompt tun- ing,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 8, 2025, pp. 8718–8726

    2025

  19. [26]

    Learning with enriched inductive biases for vision-language models,

    L. Yang, R.-Y . Zhang, Q. Chen, and X. Xie, “Learning with enriched inductive biases for vision-language models,” International Journal of Computer Vision, pp. 1–16, 2025

    2025

  20. [27]

    Hegraphadapter: Tuning multi-modal vision-language models with heterogeneous graph adapter,

    Y . Zhao, B. Jiang, X. Wang, Q. Xu, and J. Tang, “Hegraphadapter: Tuning multi-modal vision-language models with heterogeneous graph adapter,” arXiv preprint arXiv:2410.07854 , 2024

    2024 arXiv

  21. [28]

    Language models 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 models are few-shot learners,” NeurIPS, vol. 33, 2020

    2020

  22. [29]

    Amu-tuning: Effective logit bias for clip-based few-shot learning,

    Y . Tang, Z. Lin, Q. Wang, P. Zhu, and Q. Hu, “Amu-tuning: Effective logit bias for clip-based few-shot learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 323–23 333

    2024

  23. [30]

    Bayesian exploration of pre- trained models for low-shot image classification,

    Y . Miao, Y . Lei, F. Zhou, and Z. Deng, “Bayesian exploration of pre- trained models for low-shot image classification,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 849–23 859

    2024

  24. [31]

    An empirical study of training self- supervised vision transformers,

    X. Chen, S. Xie, and K. He, “An empirical study of training self- supervised vision transformers,” in ICCV, October 2021, pp. 9640–9649

    2021

  25. [32]

    Emerging properties in self-supervised vision transformers,

    M. Caron, H. Touvron, I. Misra, H. J ´egou, J. Mairal, P. Bojanowski, and A. Joulin, “Emerging properties in self-supervised vision transformers,” in ICCV, 2021, pp. 9650–9660. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12

    2021

  26. [33]

    Confidence estimation of classification based on the distribution of the neural network output layer,

    A. A. Taha, L. Hennig, and P. Knoth, “Confidence estimation of classification based on the distribution of the neural network output layer,” arXiv preprint arXiv:2210.07745 , 2022

    2022 arXiv

  27. [34]

    Zero-shot text-to-image generation,

    A. Ramesh, M. Pavlov, G. Goh, S. Gray, C. V oss, A. Radford, M. Chen, and I. Sutskever, “Zero-shot text-to-image generation,” in ICML. PMLR, 2021, pp. 8821–8831

    2021

  28. [35]

    Distribution-aware prompt tuning for vision-language models,

    E. Cho, J. Kim, and H. J. Kim, “Distribution-aware prompt tuning for vision-language models,” in Proceedings of the IEEE/CVF international conference on computer vision , 2023, pp. 22 004–22 013

    2023

  29. [36]

    Make prompts adaptable: Bayesian modeling for vision-language prompt learning with data-dependent prior,

    Y . Cho, H. Bae, S. Shin, Y . D. Youn, W. Joo, and I.-C. Moon, “Make prompts adaptable: Bayesian modeling for vision-language prompt learning with data-dependent prior,” in Proceedings of the AAAI Confer- ence on Artificial Intelligence , vol. 38, no. 10, 2024, pp. 11 552–11 560

    2024

  30. [37]

    Any- shift prompting for generalization over distributions,

    Z. Xiao, J. Shen, M. M. Derakhshani, S. Liao, and C. G. Snoek, “Any- shift prompting for generalization over distributions,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 13 849–13 860

    2024

  31. [38]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” in ICLR, 2017

    2017

  32. [39]

    Graph attention networks,

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Li `o, and Y . Bengio, “Graph attention networks,” in ICLR, 2018

    2018

  33. [40]

    Label propagation for zero-shot classifi- cation with vision-language models,

    Y . Kalantidis, G. Tolias et al., “Label propagation for zero-shot classifi- cation with vision-language models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 209–23 218

    2024

  34. [41]

    Efficient and context- aware label propagation for zero-/few-shot training-free adaptation of vision-language model,

    Y . Li, Y . Su, A. Goodge, K. Jia, and X. Xu, “Efficient and context- aware label propagation for zero-/few-shot training-free adaptation of vision-language model,” in ICLR, 2025

    2025

  35. [42]

    On vertex, edge, and vertex-edge random graphs,

    E. Beer, J. Allen Fill, S. Janson, and E. Scheinerman, “On vertex, edge, and vertex-edge random graphs,” in 2011 Proceedings of the Eighth Workshop on Analytic Algorithmics and Combinatorics (ANALCO) . SIAM, 2011, pp. 16–22

    2011

  36. [43]

    Awt: Transferring vision- language models via augmentation, weighting, and transportation,

    Y . Zhu, Y . Ji, Z. Zhao, G. Wu, and L. Wang, “Awt: Transferring vision- language models via augmentation, weighting, and transportation,” Ad- vances in Neural Information Processing Systems , vol. 37, 2024

    2024

  37. [45]

    Auto-encoding variational bayes,

    D. P. Kingma, M. Welling et al., “Auto-encoding variational bayes,” in ICLR, 2014

    2014

  38. [46]

    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,” in CVPR. Ieee, 2009, pp. 248–255

    2009

  39. [47]

    3D object representations for fine-grained categorization,

    J. Krause, M. Stark, J. Deng, and L. Fei-Fei, “3D object representations for fine-grained categorization,” in ICCV workshops, 2013, pp. 554–561

    2013

  40. [48]

    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,” in CVPR workshop. IEEE, 2004, pp. 178–178

    2004

  41. [49]

    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,” arXiv preprint arXiv:1212.0402, 2012

    2012 arXiv

  42. [50]

    Automated flower classification over a large number of classes,

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

    2008

  43. [51]

    Food-101–Mining discriminative components with random forests,

    L. Bossard, M. Guillaumin, and L. Van Gool, “Food-101–Mining discriminative components with random forests,” in ECCV. Springer, 2014, pp. 446–461

    2014

  44. [52]

    Describing textures in the wild,

    M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, and A. Vedaldi, “Describing textures in the wild,” in CVPR, 2014, pp. 3606–3613

    2014

  45. [53]

    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

  46. [54]

    Fine- grained visual classification of aircraft,

    S. Maji, E. Rahtu, J. Kannala, M. Blaschko, and A. Vedaldi, “Fine- grained visual classification of aircraft,” arXiv preprint arXiv:1306.5151, 2013

    2013 arXiv

  47. [55]

    Cats and dogs,

    O. M. Parkhi, A. Vedaldi, A. Zisserman, and C. V . Jawahar, “Cats and dogs,” in CVPR. IEEE Computer Society, 2012, pp. 3498–3505

    2012

  48. [56]

    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,” in CVPR. IEEE, 2010, pp. 3485–3492

    2010

  49. [57]

    Do ImageNet classifiers generalize to ImageNet?

    B. Recht, R. Roelofs, L. Schmidt, and V . Shankar, “Do ImageNet classifiers generalize to ImageNet?” in ICML. 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,” NeurIPS, vol. 32, 2019

    2019

  51. [59]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” in ICLR, 2019

    2019

  52. [60]

    Fast and accurate deep network learning by exponential linear units (elus),

    D.-A. Clevert, T. Unterthiner, and S. Hochreiter, “Fast and accurate deep network learning by exponential linear units (elus),” in ICLR, 2016

    2016

  53. [61]

    Deep sparse rectifier neural networks,

    X. Glorot, A. Bordes, and Y . Bengio, “Deep sparse rectifier neural networks,” in Proceedings of the fourteenth international conference on artificial intelligence and statistics . JMLR Workshop and Conference Proceedings, 2011, pp. 315–323

    2011

  54. [62]

    Dual memory networks: A versatile adaptation approach for vision-language models,

    Y . Zhang, W. Zhu, H. Tang, Z. Ma, K. Zhou, and L. Zhang, “Dual memory networks: A versatile adaptation approach for vision-language models,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2024, pp. 28 718–28 728

    2024

  55. [63]

    Self-regulating prompts: Foundational model adaptation without forgetting,

    M. U. Khattak, S. T. Wasim, M. Naseer, S. Khan, M.-H. Yang, and F. S. Khan, “Self-regulating prompts: Foundational model adaptation without forgetting,” in Proceedings of the IEEE/CVF international conference on computer vision , 2023, pp. 15 190–15 200

    2023

  56. [64]

    Mmrl: Multi-modal representation learning for vision-language models,

    Y . Guo and X. Gu, “Mmrl: Multi-modal representation learning for vision-language models,” in Proceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 25 015–25 025

    2025

  57. [65]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in CVPR, 2016, pp. 770–778

    2016

  58. [66]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” in International Conference on Learning Representations , 2020

    2020

  59. [67]

    Visualizing data using t-sne

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.” Journal of machine learning research , vol. 9, no. 11, 2008

    2008

Pith tools

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