Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Accelerating Conditional Prompt Learning via Masked Image Modeling for Vision-Language Models

T0 review · 4 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read ProMIM claims that generating prompt tokens from randomly masked image patches, rather than full images, reduces overfitting in CLIP prompt learning and improves accuracy on unseen classes with negligible added cost.

desk verdict ProMIM's masked-conditioning idea is cheap and the CoCoOp gain looks real, but the paper's own ablation shows most of the gain comes from the borrowed KgCoOp term, not the masking. read the letter →

arxiv 2508.04942 v1 pith:FFEDK6B5 submitted 2025-08-07 cs.CV

classification cs.CV
keywords promptlearningvision-languagemodelsCLIPmaskedimagemodelingfew-shotclassificationbase-to-newgeneralizationdomainparameter-efficientadaptation
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

ProMIM is a plug-and-play addition to conditional prompt learning for vision-language models such as CLIP. It claims that generating per-image prompt tokens from a randomly masked version of the image, instead of the full image, stops visual details from leaking into the text branch and making prompts overfit seen classes. The paper reports that this change improves accuracy on unseen classes across the standard 11-dataset benchmark, including when added to stronger prompt learners, while adding almost no training or inference cost. A reader should care because prompt learning is the lightweight way to adapt large vision-language models, and a drop-in module that improves generalization without new parameters would make that adaptation more practical.

What carries the argument

The central mechanism is masked-image-conditioned prompt generation. Where CoCoOp computes $\pi = h_\theta(x)$ from full image features, ProMIM computes $\pi = h_\theta(x_{\text{masked}})$ with 75% of patches randomly removed. Because only visible patches are encoded, the extra pass is cheap; because prompt tokens are built from incomplete visual evidence, the paper argues the text branch cannot memorize class-specific visual details. The auxiliary KgCoOp-style loss $L_{kg}$ is the second supporting mechanism, anchoring the soft prompts to CLIP's hand-crafted template.

What would settle it

Train ProMIM with the same loss and compute budget but replace random patch masking with a stochastic corruption that removes no information, such as additive Gaussian noise on patch embeddings or random dropout of a matched number of feature channels; if unseen-class accuracy rises by the same roughly two points, the information-removal mechanism is not what carries the result. An even sharper test holds the KgCoOp term fixed and compares 75% random masking to 75% of patches replaced by their mean patch, measuring New accuracy on the 11 benchmark datasets.

Watch

Extended reading notes

Core claim

ProMIM starts from CoCoOp-style conditional prompt learning, where a lightweight meta-network $h_\theta$ turns image features into context tokens prepended to the class-name text prompt. In ProMIM, the meta-network is fed features from a randomly masked image rather than the full image. Following masked-autoencoder practice, the frozen CLIP image encoder processes only the visible patches; at 75% masking this second forward pass costs about a quarter of a full-image pass. The final training loss is the cross-entropy classification loss plus a KgCoOp-style term $\lambda L_{kg}$ that pulls the soft-prompt text embeddings toward CLIP's hand-crafted template embeddings. On the 11-dataset base-to

Load-bearing premise

The paper's central claim depends on masking doing something specific—cutting off visual details that let prompts memorize training classes—rather than just adding random noise that happens to help generalization.

Editorial extensions

If this is right

  • Plugging ProMIM into CoCoOp lifts New-class accuracy by about 2.3 points averaged over 11 datasets, while Base accuracy drops only slightly, so standard conditional prompt learners can become more transferable without architectural changes.
  • Adding ProMIM to DePT and TCP also improves their harmonic means (e.g., TCP H from 79.51 to 80.01), indicating the benefit is not tied to one specific meta-network design.
  • At 75% masking, the extra image-encoder pass costs roughly a quarter of a full-image forward pass, and the reported training overhead is a few minutes, making the method viable when compute is limited.
  • Simple random patch masking works better than block masking, especially at high ratios, giving a concrete design constraint for anyone reusing ProMIM.

Reading between the lines

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

  • The 'data leakage' explanation is not uniquely established by the experiments: random patch masking could equally act as a strong input-level regularizer for the meta-network, and the same gains might come from any high-variance stochastic corruption of the visual features. A direct comparison against feature dropout would settle this.
  • Because 75% masking both improves accuracy and reduces the cost of the extra forward pass, efficiency and accuracy gains are entangled; a version that uses full-image features but matches the compute budget (e.g., lower resolution) would test whether information removal itself matters.
  • The largest gains appear on hard, fine-grained datasets (FGVC Aircraft New accuracy +9.92, EuroSAT +7.99), so ProMIM may be especially helpful when base classes share spurious background cues; users on such datasets can expect more than the two-point average improvement.
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 ProMIM, a plug-and-play enhancement for conditional prompt learning in vision-language models. The method randomly masks a large fraction of image patches before the frozen CLIP image encoder and uses the resulting masked-image features as input to a meta-network that generates instance-conditioned prompts. A KgCoOp-style text-side regularization term is added to the training objective. Experiments on 11 datasets cover base-to-new generalization, cross-dataset transfer, and domain generalization, with ViT-B/16 backbones. The paper reports average improvements over CoCoOp, DePT, and TCP, with negligible additional training time, inference time, and learnable parameters.

Significance. If the central claim held, ProMIM would be a simple and broadly applicable plug-in: any CoCoOp-style prompt learner could gain ~1-2 points on unseen classes with no new parameters and almost no compute overhead. The paper is clearly written and the negligible-compute claim is well supported by Table 6(a). However, the evidence for the specific MIM mechanism is currently weak. The ablation in Table 5 shows that the marginal contribution of masking on top of the KgCoOp loss is only +0.16 H, the 'consistently boosts' claim is contradicted by several per-dataset drops, and Section 5.4 contains an internal contradiction about the optimal mask ratio. If these issues are resolved with additional isolation experiments and corrected claims, the contribution could be a useful empirical addition to the prompt-learning literature; at present the significance is moderate.

major comments (4)
  1. [§5.2/Table 5] The MIM-specific contribution is not isolated and is small. CoCoOp H=75.83, KgCoOp-only H=77.00, MIM-only H=76.15, and full ProMIM H=77.16. Thus adding MIM to KgCoOp gives +0.16 H and adding MIM to CoCoOp gives +0.32 H, both likely within run-to-run variation. No ablation compares masked-image conditioning to generic stochastic augmentation (e.g., random patch dropout or random crop) at matched compute, so the claimed 'data leakage' mechanism is untested. Please add such an ablation and report error bars.
  2. [§5.4/Table 7] The text states 'the best results are observed when 95% of the input image is masked,' but Table 7 reports H=75.36 for mask ratio 0.95 and H=77.16 for 0.75. This is a direct internal contradiction and undercuts the recommended hyperparameter. The following sentence, 'when a significant portion of the input image is conditioned, performance declines,' also appears to confuse masked and conditioned. Please correct and re-analyze the mask-ratio selection.
  3. [Abstract/§4.1/Tables 1-2] The claim that ProMIM 'consistently boosts generalization performance' is contradicted by per-dataset results. For example, in Table 1, DePT+Pets H drops from 96.46 to 95.78 and TCP+DTD H drops from 68.25 to 66.94. In Table 2, DePT+ProMIM on Flowers drops from 72.17 to 69.97. The average gains are positive, but 'consistently' is not supported. Please qualify the claim to 'on average improves' or explain these exceptions.
  4. [§4 Baselines/Table 1] Baseline numbers are 'directly sourced from their respective original publications' rather than re-run under the same codebase. Since many reported gains are 0.1-0.4 H, cross-publication comparisons without shared code or standard deviations are insufficient to support a plug-and-play claim. Please re-run the baselines in the authors' environment or provide error bars and significance statements for all key tables.
minor comments (6)
  1. [Figure 1 caption] The caption says 'six prompt tuning methods' but the figure lists seven: VPT, CoCoOp, KgCoOp, MaPLe, PromptSRC, DePT, and TCP. Please correct.
  2. [§5.6/Figure 4] The text says 'a consistent improvement in accuracy on 10 out of 11 datasets,' but Figure 4(a) shows positive gains on all 11 datasets. Also, the text says 'more than 10% improvement on FGVCAircraft,' while the bar shows +9.92%. Please reconcile.
  3. [§3.2 (Eq. 4)] Equation (4) uses x for the full image in the similarity computation and x_masked for prompt generation. This is clear from context, but the notation should be defined explicitly to avoid confusion about whether the masked features or full features are used for classification.
  4. [Table 6(a)] The learnable-parameter count for CoOp is listed as 8K. With 4 context tokens of 512 dimensions each, the expected count is 2K (2048). Please verify the parameter counts for all methods.
  5. [§4 Implementation details] The text reports training time on a single H100 GPU 'within a day for all 11 datasets.' Table 6(a) gives CoCoOp 420 minutes, which is consistent, but the phrase should be made more precise (e.g., total vs. per-dataset time).
  6. [§1 and throughout] The phrase 'masking only visible image patches' is confusing; the method masks a subset of patches and keeps the visible ones. Please rephrase to 'masking image patches and using the remaining visible patches.'

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: ProMIM is an empirical modification evaluated on external benchmarks; the only mild concern is hyperparameter selection on the same average used for the headline, which is a selection-on-test issue, not a reduction by construction.

full rationale

The paper makes no formal derivation claim; ProMIM is defined by Eq. 4-6, where masked image features feed the meta-net and the objective is cross-entropy plus the borrowed KgCoOp regularizer. No predicted quantity is defined in terms of the method's own fitted outputs, and no load-bearing result is imported from a self-citation: reference [4] is unrelated to prompt learning, and KgCoOp [41] is external prior work. The central evidence is empirical comparison on 11 established datasets. The closest to a circularity concern is that lambda=2.0 and mask ratio=0.75 were selected using the same 11-dataset H average that is then reported as the headline result (Section 5.4, Tables 7-8), which is selection on the evaluation metric; however, this does not make the reported accuracy equal to the tuning criterion by construction, and it is a standard hyperparameter choice rather than a fitted parameter renamed as a prediction. The MIM-specific contribution is small (+0.16 H over KgCoOp in Table 5) and the data-leakage mechanism is not isolated from generic augmentation, but these are evidentiary weaknesses, not circularity. The text's claim that 95% masking is best contradicts Table 7 (0.75 best), which is an internal inconsistency rather than a circular step.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method introduces no new entities. It depends on the empirical assumption that masked CLIP features reduce overfitting, on the prior KgCoOp regularization, and on comparability of baselines from original publications. Two hyperparameters, the mask ratio and lambda, are tuned on the same benchmarks used for the headline results.

free parameters (3)
  • L_kg weight lambda = 2.0
    Selected in Section 5.4, Table 8 as the value maximizing the 11-dataset harmonic mean used in every headline table.
  • mask ratio = 0.75
    Selected in Section 5.4, Table 7 as the best value (H=77.16); the text in Section 5.4 says 0.95 is best, contradicting the table.
  • mask sampling strategy = random
    Selected in Section 5.5, Table 6b; random beats block-wise masking in the authors' chosen meta-evaluation.
assumptions (4)
  • domain assumption Randomly masked CLIP features are a better conditioning signal than full image features
    Section 3.2, central mechanism; no control isolates masking from generic regularization.
  • domain assumption KgCoOp's L_kg regularization improves unseen-class generalization when added to ProMIM
    Equation 5, Section 3.2; adopted from prior work and kept because it helped in the authors' ablations.
  • domain assumption The pretrained CLIP image encoder produces usable features from partially masked images without architectural changes
    Section 3.2; required for the masked feature path to work, following MAE-style patch processing.
  • domain assumption Published baseline numbers are comparable to the authors' own runs
    Implementation details; no shared harness, and the DePT domain generalization baseline is missing in Table 3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Accelerating Conditional Prompt Learning via Masked Image Modeling for Vision-Language Models." pith.science (2026). https://pith.science/paper/FFEDK6B5

@misc{pith2026250804942,
  author       = {Pith},
  title        = {Pith review of: Accelerating Conditional Prompt Learning via Masked Image Modeling for Vision-Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FFEDK6B5}},
  note         = {Machine review of arXiv:2508.04942}
}
read the original abstract

Vision-language models (VLMs) like CLIP excel in zero-shot learning but often require resource-intensive training to adapt to new tasks. Prompt learning techniques, such as CoOp and CoCoOp, offer efficient adaptation but tend to overfit to known classes, limiting generalization to unseen categories. We introduce ProMIM, a plug-and-play framework that enhances conditional prompt learning by integrating masked image modeling (MIM) into existing VLM pipelines. ProMIM leverages a simple yet effective masking strategy to generate robust, instance-conditioned prompts, seamlessly augmenting methods like CoOp and CoCoOp without altering their core architectures. By masking only visible image patches and using these representations to guide prompt generation, ProMIM improves feature robustness and mitigates overfitting, all while introducing negligible additional computational cost. Extensive experiments across zero-shot and few-shot classification tasks demonstrate that ProMIM consistently boosts generalization performance when plugged into existing approaches, providing a practical, lightweight solution for real-world vision-language applications.

Figures

Figures reproduced from arXiv: 2508.04942 by the authors.

Figure 1
Figure 1. Classification accuracies of six prompt tuning meth [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Performance comparison on base-to-novel general [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The framework of the proposed ProMIM. The masked input image embedding is used to generate the learnable meta [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Extensive evaluation of ProMIM versus CoCoOp in [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Attn-Adapter: Attention Is All You Need for Online Few-shot Learner of Vision-Language Model

    cs.CV 2025-09 conditional novelty 6.0 of 10

    A dual attention adapter using support-image memory and local-global feature mixing improves CLIP few-shot and domain-shift classification.

Reference graph

Works this paper leans on

52 extracted references · 28 canonical work pages · cited by 1 Pith paper

  1. [1]

    Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katherine Millican, Malcolm Reynolds, et al

  2. [2]

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. 2021. Beit: Bert pre-training of image transformers. arXiv preprint arXiv:2106.08254 (2021)

  3. [3]

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. 2014. Food-101–mining discriminative components with random forests. In Computer vision–ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, proceedings, part VI 13. Springer, 446–461

  4. [4]

    Phuoc-Nguyen Bui, Duc-Tai Le, and Hyunseung Choo. 2024. Visual-textual matching attention for lesion segmentation in chest images. In International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, 702–711

  5. [5]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. In Interna- tional conference on machine learning . PMLR, 1597–1607

  6. [6]

    Jaemin Cho, Jie Lei, Hao Tan, and Mohit Bansal. 2021. Unifying vision-and- language tasks via text generation. In International Conference on Machine Learn- ing. PMLR, 1931–1942

  7. [7]

    Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and An- drea Vedaldi. 2014. Describing textures in the wild. In Proceedings of the IEEE conference on computer vision and pattern recognition . 3606–3613

  8. [8]

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. 2009. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition . Ieee, 248–255

Show all 52 references
  1. [9]

    Alexey Dosovitskiy. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929 (2020)

  2. [10]

    Li Fei-Fei, Rob Fergus, and Pietro Perona. 2004. Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories. In 2004 conference on computer vision and pattern recognition workshop. IEEE, 178–178

  3. [11]

    Peng Gao, Shijie Geng, Renrui Zhang, Teli Ma, Rongyao Fang, Yongfeng Zhang, Hongsheng Li, and Yu Qiao. 2024. Clip-adapter: Better vision-language models with feature adapters. International Journal of Computer Vision 132, 2 (2024), 581–595

  4. [12]

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick

  5. [13]

    Patrick Helber, Benjamin Bischke, Andreas Dengel, and Damian Borth. 2019. Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 12, 7 (2019), 2217–2226

  6. [14]

    In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 16000–16009

  7. [15]

    Dan Hendrycks, Kevin Zhao, Steven Basart, Jacob Steinhardt, and Dawn Song

  8. [16]

    Dan Hendrycks, Steven Basart, Norman Mu, Saurav Kadavath, Frank Wang, Evan Dorundo, Rahul Desai, Tyler Zhu, Samyak Parajuli, Mike Guo, et al. 2021. The many faces of robustness: A critical analysis of out-of-distribution generalization. In Proceedings of the IEEE/CVF internati...

  9. [17]

    Muhammad Uzair Khattak, Hanoona Rasheed, Muhammad Maaz, Salman Khan, and Fahad Shahbaz Khan. 2023. Maple: Multi-modal prompt learning. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 19113–19122

  10. [18]

    Muhammad Uzair Khattak, Syed Talal Wasim, Muzammal Naseer, Salman Khan, Ming-Hsuan Yang, and Fahad Shahbaz Khan. 2023. Self-regulating prompts: Foundational model adaptation without forgetting. InProceedings of the IEEE/CVF International Conference on Computer Vision . 15190–15200

  11. [19]

    Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc Le, Yun-Hsuan Sung, Zhen Li, and Tom Duerig. 2021. Scaling up visual and vision- language representation learning with noisy text supervision. In International conference on machine learning . PMLR, 4904–4916

  12. [20]

    Xiangwen Kong and Xiangyu Zhang. 2023. Understanding masked image mod- eling via learning occlusion invariant feature. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 6241–6251

  13. [21]

    Jonathan Krause, Michael Stark, Jia Deng, and Li Fei-Fei. 2013. 3d object repre- sentations for fine-grained categorization. In Proceedings of the IEEE international conference on computer vision workshops . 554–561

  14. [22]

    Wonjae Kim, Bokyung Son, and Ildoo Kim. 2021. Vilt: Vision-and-language trans- former without convolution or region supervision. In International conference on machine learning. PMLR, 5583–5594

  15. [23]

    Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. 2023. Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. Comput. Surveys 55, 9 (2023), 1–35

  16. [24]

    Jiasen Lu, Dhruv Batra, Devi Parikh, and Stefan Lee. 2019. Vilbert: Pretrain- ing task-agnostic visiolinguistic representations for vision-and-language tasks. Advances in neural information processing systems 32 (2019)

  17. [25]

    Siting Li, Pang Wei Koh, and Simon Shaolei Du. 2024. On Erroneous Agreements of CLIP Image Embeddings. arXiv:2411.05195 [cs.LG] https://arxiv.org/abs/2411. 05195

  18. [26]

    Subhransu Maji, Esa Rahtu, Juho Kannala, Matthew Blaschko, and Andrea Vedaldi

  19. [27]

    Maria-Elena Nilsback and Andrew Zisserman. 2008. Automated flower classifica- tion over a large number of classes. In 2008 Sixth Indian conference on computer vision, graphics & image processing . IEEE, 722–729

  20. [28]

    Chengcheng Ma, Yang Liu, Jiankang Deng, Lingxi Xie, Weiming Dong, and Changsheng Xu. 2023. Understanding and Mitigating Overfitting in Prompt Tuning for Vision-Language Models. IEEE Transactions on Circuits and Systems for Video Technology 33, 9 (2023), 4616–4629. doi:10.1109/...

  21. [29]

    Fabio Petroni, Tim Rocktäschel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, Alexander H Miller, and Sebastian Riedel. 2019. Language models as knowledge bases? arXiv preprint arXiv:1909.01066 (2019)

  22. [30]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learni...

  23. [31]

    Yongming Rao, Wenliang Zhao, Guangyi Chen, Yansong Tang, Zheng Zhu, Guan Huang, Jie Zhou, and Jiwen Lu. 2022. Denseclip: Language-guided dense predic- tion with context-aware prompting. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 18082–18091

  24. [32]

    Omkar M Parkhi, Andrea Vedaldi, Andrew Zisserman, and CV Jawahar. 2012. Cats and dogs. In 2012 IEEE conference on computer vision and pattern recognition . IEEE, 3498–3505

  25. [33]

    Sheng Shen, Shijia Yang, Tianjun Zhang, Bohan Zhai, Joseph E Gonzalez, Kurt Keutzer, and Trevor Darrell. 2024. Multitask vision-language prompt tuning. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision . 5656–5667

  26. [34]

    Lingxue Song, Dihong Gong, Zhifeng Li, Changsong Liu, and Wei Liu. 2019. Occlusion robust face recognition based on mask learning with pairwise differ- ential siamese network. In Proceedings of the IEEE/CVF international conference on computer vision. 773–782. LAVA ’25, Octobe...

  27. [35]

    Khurram Soomro, Amir Roshan Zamir, and Mubarak Shah. 2012. UCF101: A dataset of 101 human actions classes from videos in the wild. arXiv preprint arXiv:1212.0402 (2012)

  28. [36]

    Benjamin Recht, Rebecca Roelofs, Ludwig Schmidt, and Vaishaal Shankar. 2019. Do imagenet classifiers generalize to imagenet?. In International conference on machine learning. PMLR, 5389–5400

  29. [37]

    Haohan Wang, Songwei Ge, Zachary Lipton, and Eric P Xing. 2019. Learning robust global representations by penalizing local predictive power. Advances in Neural Information Processing Systems 32 (2019)

  30. [38]

    Zirui Wang, Jiahui Yu, Adams Wei Yu, Zihang Dai, Yulia Tsvetkov, and Yuan Cao

  31. [39]

    Ge Wu, Xin Zhang, Zheng Li, Zhaowei Chen, Jiajun Liang, Jian Yang, and Xiang Li. 2024. Cascade prompt learning for vision-language model adaptation. In European Conference on Computer Vision . Springer, 304–321

  32. [40]

    Maria Tsimpoukelli, Jacob L Menick, Serkan Cabi, SM Eslami, Oriol Vinyals, and Felix Hill. 2021. Multimodal few-shot learning with frozen language models. Advances in Neural Information Processing Systems 34 (2021), 200–212

  33. [41]

    Hantao Yao, Rui Zhang, and Changsheng Xu. 2023. Visual-language prompt tun- ing with knowledge-guided context optimization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 6757–6767

  34. [42]

    Hantao Yao, Rui Zhang, and Changsheng Xu. 2024. TCP: Textual-based Class- aware Prompt tuning for Visual-Language Model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 23438–23448

  35. [43]

    arXiv preprint arXiv:2108.10904 (2021)

    Simvlm: Simple visual language model pretraining with weak supervision. arXiv preprint arXiv:2108.10904 (2021)

  36. [44]

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. 2022. Conditional prompt learning for vision-language models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 16816–16825

  37. [45]

    Jianxiong Xiao, James Hays, Krista A Ehinger, Aude Oliva, and Antonio Torralba

  38. [46]

    Beier Zhu, Yulei Niu, Yucheng Han, Yue Wu, and Hanwang Zhang. 2023. Prompt- aligned gradient for prompt tuning. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 15659–15669

  39. [49]

    Ji Zhang, Shihan Wu, Lianli Gao, Heng Tao Shen, and Jingkuan Song. 2024. Dept: Decoupled prompt tuning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 12924–12933

  40. [51]

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. 2022. Learning to prompt for vision-language models. International Journal of Computer Vision 130, 9 (2022), 2337–2348

  41. [2010]

    In 2010 IEEE computer society conference on computer vision and pattern recognition

    Sun database: Large-scale scene recognition from abbey to zoo. In 2010 IEEE computer society conference on computer vision and pattern recognition . IEEE, 3485–3492

  42. [2013]

    arXiv preprint arXiv:1306.5151 (2013)

    Fine-grained visual classification of aircraft. arXiv preprint arXiv:1306.5151 (2013)

  43. [2021]

    In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Natural adversarial examples. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 15262–15271

  44. [2022]

    Advances in neural information processing systems 35 (2022), 23716–23736

    Flamingo: a visual language model for few-shot learning. Advances in neural information processing systems 35 (2022), 23716–23736

Pith tools

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