Pith. sign in

REVIEW 5 major objections 5 minor 52 references

UCDR-Adapter: Exploring Adaptation of Pre-Trained Vision-Language Models for Universal Cross-Domain Retrieval

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read UCDR-Adapter claims that dynamic prompts, synthesized by attending over masked source prompts, let a frozen CLIP model retrieve across unseen domains and classes more accurately than static-prompt baselines.

desk verdict A reasonably built parameter-efficient prompt-tuning variant for UCDR, but the 'consistently outperforms' headline doesn't survive contact with its own tables. read the letter →

arxiv 2412.10680 v1 pith:5IA4R3A7 submitted 2024-12-14 cs.CV cs.IRcs.MM

Haoyu Jiang , Zhi-Qi Cheng , Gabriel Moreira , Jiawen Zhu , Jingdong Sun , Bukun Ren , Jun-Yan He , Qi Dai
show 1 more author
Xian-Sheng Hua
This is my paper · ORCID
classification cs.CVcs.IRcs.MM
keywords universalcross-domainretrievalprompttuningvision-languagemodelsCLIPdynamicgenerationadaptermodulesimagemomentumcontrast
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

The paper introduces UCDR-Adapter, a two-phase method to adapt a frozen CLIP model for Universal Cross-Domain Retrieval (UCDR), where test queries can come from unseen domains and classes. Phase one learns class- and domain-specific prompts under momentum updates, triplet loss, and image-text contrastive loss; phase two generates dynamic target prompts for novel inputs by attending over masked source prompts. The central claim is that this dynamic prompt generation outperforms the static-prompt ProS baseline and earlier methods on the UCDR, UdCDR, and UcCDR benchmarks, while using only the image branch at inference and adding 2.36M trainable parameters. Success would make cross-domain retrieval practical in settings where new categories and visual domains appear after deployment, without needing text descriptions at query time.

What carries the argument

The load-bearing object is the Target Prompt Generation (TPG) module. Given masked source prompts $U' = U \circ (1-\delta_d)$ and $V' = V \circ (1-\delta_c)$, it computes attention weights $w_d = \text{Attn}(I_g, U')$ and $w_c = \text{Attn}(I_g, V')$, then forms target prompts as weighted combinations of the masked rows (the paper's Eqs. 5-8). These are concatenated, projected to the encoder input dimension, and added to the image features of a frozen CLIP ViT-B/32 encoder. The same module serves all three retrieval subtasks without modification, and it is what distinguishes UCDR-Adapter from static-prompt methods like ProS.

What would settle it

The claim that UCDR-Adapter 'consistently outperforms ProS in most cases' is tested directly by the paper's Tables 2 and 4: on the Quickdraw holdout ProS achieves 0.2842 mAP@200 against 0.2794 for UCDR-Adapter, and on TU-Berlin ProS achieves 0.6675 mAP@All against 0.6581. An additional experiment that would localize the cause is to substitute uniform attention weights for the learned weights in Eqs. 7-8; if retrieval accuracy is unchanged, the attention weighting itself is not the source of the gains.

Watch

Extended reading notes

Core claim

The paper claims that adapting a pre-trained vision-language model to unseen domains and classes in retrieval is better done with dynamic, image-conditioned prompts than with static prompts. The mechanism is a Target Prompt Generation (TPG) module that masks the rows of the learned domain prompt matrix U and class prompt matrix V, computes soft-attention weights between the input image features and the masked rows, and forms target prompts as the weighted combinations of those masked rows. These generated prompts are injected into the frozen image encoder, and the resulting features are aligned to text embeddings through the contrastive loss from phase one. At test time, the text encoder and all textual inputs are discarded, so retrieval uses only the image branch. The authors report consistent improvements over the static-prompt ProS method 'in most cases' and larger gains over SASA, with the full system reaching 0.6591 mAP@200 on the DomainNet UCDR Sketch holdout versus 0.6457 for ProS.

Load-bearing premise

The premise that carries the method is that prompts useful for an unseen class or domain can be synthesized as a soft-attention-weighted combination of the masked rows of the source prompt matrices; if the masked rows do not span a space containing useful prompts for novel inputs, the Target Prompt Generation stage contributes nothing beyond the phase-one prompts.

Editorial extensions

If this is right

  • If the central claim is right, a frozen CLIP model with 2.36M trainable parameters can serve retrieval when both domain and class are unseen, eliminating test-time text input entirely.
  • The TPG masking strategy becomes a general recipe for synthesizing prompts for novel categories and domains from a fixed source-prompt bank, applicable beyond retrieval to classification or detection.
  • The reported UcCDR gains on Sketchy and TU-Berlin suggest the method transfers to fine-grained, sketch-based retrieval and would apply to other sketch-photo benchmarks.
  • The momentum queues and triplet loss are claimed to improve inter-class separation for unseen classes, implying the adapted features are more linearly separable than raw CLIP features.

Reading between the lines

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

  • The largest ablation gain comes from adding the two-phase TPG schedule (+6.82% mAP over one-phase training), while isolating the masking itself gains only +0.66% (Table 5); this suggests the benefit may come more from the two-phase optimization than from the attention-weighted masking, a distinction the paper does not make.
  • A natural extension, untested in the paper, is whether the same mask-and-attend prompt synthesis transfers to very different visual distributions such as medical or satellite imagery; the current benchmarks are all natural-image domains.
  • Because the paper reports single-run results without variance or significance tests, the margins under one mAP point separating UCDR-Adapter from ProS on several configurations would need repeated-run evidence to support the claim of consistent superiority.
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

5 major / 5 minor

Summary. The paper proposes UCDR-Adapter, a method for universal cross-domain retrieval (UCDR) that adapts a frozen CLIP model by learning domain- and class-specific prompt matrices in a first phase and then generating target prompts for unseen domains/classes through soft attention over masked source prompts in a second phase. Training uses a triplet loss with momentum-updated queues and an image-text contrastive loss with a learnable textual semantic template. At test time, only the image branch and the generated prompts are used. Experiments are reported on DomainNet, Sketchy, and TU-Berlin for UCDR, UdCDR, and UcCDR tasks, comparing with SnMpNet, SCNNet, SASA, Zero-Shot CLIP, and ProS. The paper claims superior performance, with an ablation study in Table 5 and a parameter count of 2.36M trainable parameters in Table 6.

Significance. The proposed idea of synthesizing target prompts by attending to masked source prompts is a plausible and low-cost approach to handling unseen domains and classes, and the two-phase design with momentum queues is well motivated. If the empirical claims were robustly established, the method would be a useful parameter-efficient baseline for UCDR. However, the evidence presented is mixed: the method loses to ProS on several configurations, and no significance testing or variance analysis is given. The paper also contains an apparent typo in the core TPG equations (Eqs. 7-8) and does not describe an explicit adapter module despite the title. The conceptual contribution is still interesting, and the ablation table (Table 5) provides some support for the design choices, but the central performance claim is not fully supported in the current form.

major comments (5)
  1. [Section 4.2, Eqs. (7)-(8)] The target prompt computation as written swaps the matrices. With wd = Attn(Ig, U') and wc = Attn(Ig, V'), the weighted combinations should be Pd = U' wd and Pc = V' wc. As written, Pd = V' wd and Pc = U' wc are dimensionally incompatible because U' has |Dtr| rows and V' has |Ctr| rows. Please correct the equations and clarify the intended dimensions of the attention weights and the generated prompts.
  2. [Tables 2-4 and Abstract] The claim of 'consistently outperforms ProS in most cases' is not supported by the reported numbers. ProS is better on Quickdraw and Infograph in Table 2 (e.g., mAP@200 0.2842 vs 0.2794 for Quickdraw and 0.5798 vs 0.5714 for Infograph), on Sketch in Table 3 (mAP@200 0.7385 vs 0.7332), and on TU-Berlin in Table 4 (mAP@All 0.6675 vs 0.6581). The average gains over ProS are only about +0.19 mAP points on the UCDR table and +0.38 mAP points on the UdCDR table, and no error bars, seeds, or paired tests are reported. Please either soften the claims to match the data or add repeated runs with statistical significance tests and discuss the cases where the method underperforms.
  3. [Sections 4.1-4.3] The paper's title and Section 4 promise an adapter-based method, but no adapter architecture is actually described. The text describes only learnable prompt vectors added to image features and a learnable text template; Table 6 reports 2.36M training parameters, which is plausible for prompt tuning, but the 'adapter' module itself (e.g., a bottleneck MLP inserted in the ViT) is never specified. Please state exactly which parameters constitute the adapters and how they are inserted and trained, or rename the method to avoid misleading readers.
  4. [Section 4.3 vs Section 4.1] There is a contradiction about which components are frozen. Section 4.3 says 'Frozen Image Encoder from the pre-trained CLIP model does not require any extra modifications,' but Section 4.1 states that prompts 'are optimized via backpropagation along with the vision transformer encoder parameters.' Please clarify which parameters are trainable in each phase and whether the image encoder is frozen or fine-tuned.
  5. [Section 4, first paragraph] The text calls the pipeline a 'three-phase approach' (Source Prompt Learning, Target Prompt Generation, and test-time), while the abstract and the rest of the paper refer to a 'two-phase training strategy.' This inconsistency should be reconciled, for instance by describing the test-time procedure as a separate inference phase rather than a training phase.
minor comments (5)
  1. [Abstract and Section 1] The phrase 'consistently outperforms ProS in most cases' is self-contradictory; either report consistent wins or acknowledge that the method wins in most but not all cases.
  2. [Section 4.1] The text uses 'ternary loss' instead of 'triplet loss' in the description of Eq. (2), and the sentence 'LT ripletloss on these samples' contains a spacing typo; please correct these.
  3. [Section 5.2] The UdCDR results section discusses gains over SASA but does not mention the comparison with ProS in the text; please report the relative differences to ProS and address the mixed results explicitly.
  4. [Figure 4 caption] The caption says 'Where Pd and Pc are target prompts generated for unseen domains and classes,' but the figure does not clearly label Pd and Pc; please add labels to the figure to match the text.
  5. [Implementation Details] The paper gives a project page but no code or configuration details; consider adding a reproducibility appendix with dataset splits, hyperparameter values, and the source of the ProS baseline numbers.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: UCDR-Adapter's central claims are empirical benchmark results, not derivations from their own inputs.

full rationale

UCDR-Adapter's central claim is empirical: a two-phase training procedure plus a Target Prompt Generation module improves retrieval on DomainNet, Sketchy, and TU-Berlin. I traced the derivation chain: Phase 1 optimizes class/domain prompts U, V and a textual template using LITC and triplet losses (Eqs. 1-4); Phase 2 masks the learned prompts and synthesizes target prompts via soft attention (Eqs. 5-8); at test time only the image branch is used. There is no step in which an input is defined in terms of the output, nor any fitted parameter that is later renamed a prediction. The TPG module is trained on seen classes/domains and evaluated on held-out classes/domains, so the evaluation is not forced by construction. The main baseline ProS (Ref. [12]) is co-authored by current author Zhi-Qi Cheng, but the comparison is an external benchmark run reported in Tables 2-4, not a self-citation used to justify a premise; the method is independently implemented and the results are not derived from ProS's equations. The swapped subscripts in Eqs. (7)-(8) are a potential typo and a correctness/dimensionality concern, but they do not make the central claim equivalent to its inputs. No circular step is present.

Assumptions & free parameters 7 free parameters · 3 assumptions · 0 invented entities

The method introduces no new physical or conceptual entities. It relies on hand-chosen hyperparameters, a learned prompt/template parameterization, and the unproven TPG design assumption, with the only external support being CLIP's pretraining and standard benchmarks.

free parameters (7)
  • Domain prompt matrix U = Learned (|D_tr| x 768)
    Learned in Phase 1; a central component encoding seen-domain information.
  • Class prompt matrix V = Learned (|C_tr| x 768)
    Learned in Phase 1; encodes seen-class information.
  • Text template domain vectors v_1..v_N = Learned
    Learned vectors in the 'A photo of {class} from the {v_1,...,v_N} domain' template.
  • Momentum rate alpha = 0.001
    Hand-chosen hyperparameter for momentum prompt updates (Eq. 1).
  • Triplet margin b = 0.5
    Hand-chosen margin for triplet loss (Eq. 2).
  • ITC temperature tau = 0.07
    Hand-chosen temperature for contrastive loss (Eq. 3).
  • Queue length per class = 20
    Chosen queue size for momentum feature storage.
assumptions (3)
  • domain assumption CLIP ViT-B/32 pre-trained on web-scale image-text data provides transferable representations.
    The entire method uses a frozen CLIP; its pretraining is an unexamined input (Sec. 1, Implementation Details).
  • domain assumption The UCDR benchmark splits (DomainNet 245/55/45, Sketchy 93/11/21, etc.) are valid proxies for real-world cross-domain retrieval.
    These splits are taken from prior work without independent justification (Sec. 5.1).
  • ad hoc to paper Masked source prompts attended by image features can synthesize useful target prompts for unseen classes and domains.
    This is the core design hypothesis of the TPG module (Sec. 4.2); it is not derived or proven, only empirically evaluated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UCDR-Adapter: Exploring Adaptation of Pre-Trained Vision-Language Models for Universal Cross-Domain Retrieval." pith.science (2026). https://pith.science/paper/5IA4R3A7

@misc{pith2026241210680,
  author       = {Pith},
  title        = {Pith review of: UCDR-Adapter: Exploring Adaptation of Pre-Trained Vision-Language Models for Universal Cross-Domain Retrieval},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5IA4R3A7}},
  note         = {Machine review of arXiv:2412.10680}
}
read the original abstract

Universal Cross-Domain Retrieval (UCDR) retrieves relevant images from unseen domains and classes without semantic labels, ensuring robust generalization. Existing methods commonly employ prompt tuning with pre-trained vision-language models but are inherently limited by static prompts, reducing adaptability. We propose UCDR-Adapter, which enhances pre-trained models with adapters and dynamic prompt generation through a two-phase training strategy. First, Source Adapter Learning integrates class semantics with domain-specific visual knowledge using a Learnable Textual Semantic Template and optimizes Class and Domain Prompts via momentum updates and dual loss functions for robust alignment. Second, Target Prompt Generation creates dynamic prompts by attending to masked source prompts, enabling seamless adaptation to unseen domains and classes. Unlike prior approaches, UCDR-Adapter dynamically adapts to evolving data distributions, enhancing both flexibility and generalization. During inference, only the image branch and generated prompts are used, eliminating reliance on textual inputs for highly efficient retrieval. Extensive benchmark experiments show that UCDR-Adapter consistently outperforms ProS in most cases and other state-of-the-art methods on UCDR, U(c)CDR, and U(d)CDR settings.

Figures

Figures reproduced from arXiv: 2412.10680 by the authors.

Figure 1
Figure 1. Overview of UCDR settings. Training involves seen [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. UCDR-Adapter architecture. In Phase 1 (top), [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Target Prompt Generation process. Gray indicates that [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: The results of using UCDR-Adapter for the UCDR on DomainNet. The ‘Cloud’ class from the holdout domain as the query. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: t-SNE visualization of features for 10 random unseen [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 46 canonical work pages

  1. [1]

    Contrastive learning of semantic concepts for open-set cross-domain retrieval

    Aishwarya Agarwal, Srikrishna Karanam, Balaji Vasan Srinivasan, and Biplab Banerjee. Contrastive learning of semantic concepts for open-set cross-domain retrieval. In WACV, pages 4115–4124, 2023. 2, 7

    work page 2023

  2. [2]

    Handling class-imbalance for improved zero-shot domain generaliza- tion

    Ahmad Arfeen, Titir Dutta, and Soma Biswas. Handling class-imbalance for improved zero-shot domain generaliza- tion. In BMVC, page 728, 2022. 1, 2

    work page 2022

  3. [3]

    General- izing from several related classification tasks to a new unla- beled sample

    Gilles Blanchard, Gyemin Lee, and Clayton Scott. General- izing from several related classification tasks to a new unla- beled sample. NeurIPS, 24, 2011. 2

    work page 2011

  4. [4]

    Stylip: Multi-scale style- conditioned prompt learning for clip-based domain general- ization

    Shirsha Bose, Ankit Jha, Enrico Fini, Mainak Singha, Elisa Ricci, and Biplab Banerjee. Stylip: Multi-scale style- conditioned prompt learning for clip-based domain general- ization. In WACV, pages 5542–5552, 2024. 2

    work page 2024

  5. [5]

    A review on multi- modal zero-shot learning

    Weipeng Cao, Yuhao Wu, Yixuan Sun, Haigang Zhang, Jin Ren, Dujuan Gu, and Xingkai Wang. A review on multi- modal zero-shot learning. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 13(2):e1488, 2023. 2

    work page 2023

  6. [6]

    Uniter: Universal image-text representation learning

    Yen-Chun Chen, Linjie Li, Licheng Yu, Ahmed El Kholy, Faisal Ahmed, Zhe Gan, Yu Cheng, and Jingjing Liu. Uniter: Universal image-text representation learning. In ECCV, pages 104–120, 2020. 2

    work page 2020

  7. [7]

    Video ecommerce: Towards online video advertising

    Zhi-Qi Cheng, Yang Liu, Xiao Wu, and Xian-Sheng Hua. Video ecommerce: Towards online video advertising. In Proceedings of the 24th ACM international conference on Multimedia, pages 1365–1374, 2016. 2

    work page 2016

  8. [8]

    Video ecommerce++: Toward large scale online video adver- tising

    Zhi-Qi Cheng, Xiao Wu, Yang Liu, and Xian-Sheng Hua. Video ecommerce++: Toward large scale online video adver- tising. IEEE transactions on multimedia, 19(6):1170–1183,

    work page

Show all 52 references
  1. [9]

    Video2shop: Exact matching clothes in videos to online shopping images

    Zhi-Qi Cheng, Xiao Wu, Yang Liu, and Xian-Sheng Hua. Video2shop: Exact matching clothes in videos to online shopping images. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4048–4056,

  2. [10]

    On the selection of anchors and targets for video hyperlink- ing

    Zhi-Qi Cheng, Hao Zhang, Xiao Wu, and Chong-Wah Ngo. On the selection of anchors and targets for video hyperlink- ing. In Proceedings of the 2017 acm on international con- ference on multimedia retrieval, pages 287–293, 2017. 2

    2017

  3. [11]

    An evaluation of descriptors for large-scale image retrieval from sketched feature lines

    Mathias Eitz, Kristian Hildebrand, Tamy Boubekeur, and Marc Alexa. An evaluation of descriptors for large-scale image retrieval from sketched feature lines. Computers & Graphics, 34(5):482–498, 2010. 5

    2010

  4. [12]

    Pros: Prompting-to-simulate generalized knowledge for universal cross-domain retrieval

    Kaipeng Fang, Jingkuan Song, Lianli Gao, Pengpeng Zeng, Zhi-Qi Cheng, Xiyao Li, and Heng Tao Shen. Pros: Prompting-to-simulate generalized knowledge for universal cross-domain retrieval. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pag...

    2024

  5. [13]

    Domain adaptation via prompt learning

    Chunjiang Ge, Rui Huang, Mixue Xie, Zihang Lai, Shiji Song, Shuang Li, and Gao Huang. Domain adaptation via prompt learning. IEEE Transactions on Neural Networks and Learning Systems, 2023. 2

    2023

  6. [14]

    Improv- ing diversity with adversarially learned transformations for domain generalization

    Tejas Gokhale, Rushil Anirudh, Jayaraman J Thiagarajan, Bhavya Kailkhura, Chitta Baral, and Yezhou Yang. Improv- ing diversity with adversarially learned transformations for domain generalization. In WACV, pages 434–443, 2023. 2

    2023

  7. [15]

    Mixup as locally linear out-of-manifold regularization

    Hongyu Guo, Yongyi Mao, and Richong Zhang. Mixup as locally linear out-of-manifold regularization. In AAAI, pages 3714–3722, 2019. 2

    2019

  8. [16]

    Momentum contrast for unsupervised visual rep- resentation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual rep- resentation learning. In CVPR, pages 9729–9738, 2020. 4

    2020

  9. [17]

    Vi- sual prompt tuning

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Vi- sual prompt tuning. In ECCV, pages 709–727, 2022. 2

    2022

  10. [18]

    Bert: Pre-training of deep bidirectional trans- formers for language understanding

    Jacob Devlin Ming-Wei Chang Kenton and Lee Kristina Toutanova. Bert: Pre-training of deep bidirectional trans- formers for language understanding. In NAACL-HLT, pages 4171–4186, 2019. 2

    2019

  11. [19]

    Maple: Multi-modal prompt learning

    Muhammad Uzair Khattak, Hanoona Rasheed, Muhammad Maaz, Salman Khan, and Fahad Shahbaz Khan. Maple: Multi-modal prompt learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19113–19122, 2023. 2

    2023

  12. [20]

    Self-regulating prompts: Foundational model adaptation without forgetting

    Muhammad Uzair Khattak, Syed Talal Wasim, Muzam- mal Naseer, Salman Khan, Ming-Hsuan Yang, and Fa- had Shahbaz Khan. Self-regulating prompts: Foundational model adaptation without forgetting. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 15...

    2023

  13. [21]

    Interpretability be- yond feature attribution: Quantitative testing with concept activation vectors (tcav)

    Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, et al. Interpretability be- yond feature attribution: Quantitative testing with concept activation vectors (tcav). In ICML, pages 2668–2677, 2018. 2

    2018

  14. [22]

    The power of scale for parameter-efficient prompt tuning

    Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt tuning. In EMNLP, pages 3045–3059, 2021. 2

    2021

  15. [23]

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

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. In ICML, pages 12888–12900, 2022. 1

    2022

  16. [24]

    Align before fuse: Vision and language representation learn- ing with momentum distillation

    Junnan Li, Ramprasaath Selvaraju, Akhilesh Gotmare, Shafiq Joty, Caiming Xiong, and Steven Chu Hong Hoi. Align before fuse: Vision and language representation learn- ing with momentum distillation. In NeurIPS, volume 34, pages 9694–9705, 2021. 2

    2021

  17. [25]

    Oscar: Object-semantics aligned pre-training for vision-language tasks

    Xiujun Li, Xi Yin, Chunyuan Li, Pengchuan Zhang, Xiaowei Hu, Lei Zhang, Lijuan Wang, Houdong Hu, Li Dong, Furu Wei, et al. Oscar: Object-semantics aligned pre-training for vision-language tasks. In ECCV, pages 121–137, 2020. 2

    2020

  18. [26]

    Prefix-tuning: Optimizing continuous prompts for generation

    Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimizing continuous prompts for generation. In ACL, pages 4582– 4597, 2021. 2

    2021

  19. [27]

    Deep sketch hashing: Fast free-hand sketch-based im- age retrieval

    Li Liu, Fumin Shen, Yuming Shen, Xianglong Liu, and Ling Shao. Deep sketch hashing: Fast free-hand sketch-based im- age retrieval. In CVPR, pages 2862–2871, 2017. 5

    2017

  20. [28]

    Cocoa: Context-conditional adaptation for recognizing unseen classes in unseen do- mains

    Puneet Mangla, Shivam Chandhok, Vineeth N Balasubrama- nian, and Fahad Shahbaz Khan. Cocoa: Context-conditional adaptation for recognizing unseen classes in unseen do- mains. In WACV, pages 865–874, 2022. 1, 2

    2022

  21. [29]

    Seic: Semantic embed- ding with intermediate classes for zero-shot domain general- ization

    Biswajit Mondal and Soma Biswas. Seic: Semantic embed- ding with intermediate classes for zero-shot domain general- ization. In ACCV, pages 789–806, 2022. 1, 2, 5, 6

    2022

  22. [30]

    Vireo@ trecvid 2017: Video-to-text, ad-hoc video search and video hyper- linking

    Phuong Anh Nguyen, Qing Li, Zhi-Qi Cheng, Yi-Jie Lu, Hao Zhang, Xiao Wu, and Chong-Wah Ngo. Vireo@ trecvid 2017: Video-to-text, ad-hoc video search and video hyper- linking. In TREC video retrieval evaluation, 2017. 2

    2017

  23. [31]

    Towards calibrated robust fine-tuning of vision-language models

    Changdae Oh, Hyesu Lim, Mijoo Kim, Dongyoon Han, Sangdoo Yun, Jaegul Choo, Alexander Hauptmann, Zhi-Qi Cheng, and Kyungwoo Song. Towards calibrated robust fine-tuning of vision-language models. In arXiv preprint arXiv:2311.01723, 2024. 2

    2024 arXiv

  24. [32]

    Ro- bust adaptation of foundation models with black-box visual prompting

    Changdae Oh, Gyeongdeok Seo, Geunyoung Jung, Zhi-Qi Cheng, Hosik Choi, Jiyoung Jung, and Kyungwoo Song. Ro- bust adaptation of foundation models with black-box visual prompting. In arXiv preprint arXiv:2407.17491, 2024. 2

    2024 arXiv

  25. [33]

    Universal cross-domain retrieval: Generalizing across classes and do- mains

    Soumava Paul, Titir Dutta, and Soma Biswas. Universal cross-domain retrieval: Generalizing across classes and do- mains. In ICCV, pages 12056–12064, 2021. 1, 2, 5, 6, 7, 8

    2021

  26. [34]

    Clipping: Distilling clip-based models with a student base for video- language retrieval

    Renjing Pei, Jianzhuang Liu, Weimian Li, Bin Shao, Song- cen Xu, Peng Dai, Juwei Lu, and Youliang Yan. Clipping: Distilling clip-based models with a student base for video- language retrieval. In CVPR, pages 18983–18992, 2023. 1

    2023

  27. [35]

    Moment matching for multi-source domain adaptation

    Xingchao Peng, Qinxun Bai, Xide Xia, Zijun Huang, Kate Saenko, and Bo Wang. Moment matching for multi-source domain adaptation. In ICCV, pages 1406–1415, 2019. 5

    2019

  28. [36]

    Language models as knowledge bases? In EMNLP, pages 2463–2473, 2019

    Fabio Petroni, Tim Rockt ¨aschel, Sebastian Riedel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and Alexander Miller. Language models as knowledge bases? In EMNLP, pages 2463–2473, 2019. 2

    2019

  29. [37]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICML, pages 8748–8763, 2021. 1, 2, 6

    2021

  30. [38]

    Visual semantic segmentation based on few/zero-shot learning: An overview

    Wenqi Ren, Yang Tang, Qiyu Sun, Chaoqiang Zhao, and Qing-Long Han. Visual semantic segmentation based on few/zero-shot learning: An overview. IEEE/CAA Journal of Automatica Sinica, 2023. 2

    2023

  31. [39]

    Clip for all things zero-shot sketch-based image retrieval, fine- grained or not

    Aneeshan Sain, Ayan Kumar Bhun0ia, Pinaki Nath Chowd- hury, Subhadeep Koley, Tao Xiang, and Yi-Zhe Song. Clip for all things zero-shot sketch-based image retrieval, fine- grained or not. In CVPR, pages 2765–2775, 2023. 1

    2023

  32. [40]

    The sketchy database: learning to retrieve badly drawn bunnies

    Patsorn Sangkloy, Nathan Burnell, Cusuh Ham, and James Hays. The sketchy database: learning to retrieve badly drawn bunnies. Transactions on Graphics, 35(4):1–12, 2016. 5

    2016

  33. [41]

    Structure-aware semantic-aligned network for universal cross-domain retrieval

    Jialin Tian, Xing Xu, Kai Wang, Zuo Cao, Xunliang Cai, and Heng Tao Shen. Structure-aware semantic-aligned network for universal cross-domain retrieval. In SIGIR, pages 278– 289, 2022. 1, 2, 5, 6, 7, 8

    2022

  34. [42]

    Generalizing to unseen domains: A survey on domain generalization

    Jindong Wang, Cuiling Lan, Chang Liu, Yidong Ouyang, Tao Qin, Wang Lu, Yiqiang Chen, Wenjun Zeng, and Philip Yu. Generalizing to unseen domains: A survey on domain generalization. IEEE Transactions on Knowledge and Data Engineering, 2022. 2

    2022

  35. [43]

    Image as a foreign language: Beit pretraining for vision and vision- language tasks

    Wenhui Wang, Hangbo Bao, Li Dong, Johan Bjorck, Zhil- iang Peng, Qiang Liu, Kriti Aggarwal, Owais Khan Mo- hammed, Saksham Singhal, Subhojit Som, et al. Image as a foreign language: Beit pretraining for vision and vision- language tasks. In CVPR, pages 19175–19186, 2023. 2

    2023

  36. [44]

    Ra-clip: Retrieval augmented contrastive language-image pre-training

    Chen-Wei Xie, Siyang Sun, Xiong Xiong, Yun Zheng, Deli Zhao, and Jingren Zhou. Ra-clip: Retrieval augmented contrastive language-image pre-training. In CVPR, pages 19265–19274, 2023. 2

    2023

  37. [45]

    Towards zero-shot learning: A brief review and an attention-based embedding network

    Guo-Sen Xie, Zheng Zhang, Huan Xiong, Ling Shao, and Xuelong Li. Towards zero-shot learning: A brief review and an attention-based embedding network. IEEE Transactions on Circuits and Systems for Video Technology, 33(3):1181– 1197, 2023. 2

    2023

  38. [46]

    A fourier-based framework for domain generaliza- tion

    Qinwei Xu, Ruipeng Zhang, Ya Zhang, Yanfeng Wang, and Qi Tian. A fourier-based framework for domain generaliza- tion. In CVPR, pages 14383–14392, 2021. 2

    2021

  39. [47]

    Mixup without hesitation

    Hao Yu, Huanyu Wang, and Jianxin Wu. Mixup without hesitation. In ICIG, pages 143–154, 2021. 2

    2021

  40. [48]

    Coca: Contrastive captioners are image-text foundation models

    Jiahui Yu, Zirui Wang, Vijay Vasudevan, Legg Yeung, Mo- jtaba Seyedhosseini, and Yonghui Wu. Coca: Contrastive captioners are image-text foundation models. In TMLR,

  41. [49]

    mixup: Beyond empirical risk minimiza- tion

    Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimiza- tion. In ICLR, 2018. 2

    2018

  42. [50]

    Domain generalization: A survey

    Kaiyang Zhou, Ziwei Liu, Yu Qiao, Tao Xiang, and Chen Change Loy. Domain generalization: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(04):4396–4415, 2023. 2

    2023

  43. [51]

    Conditional prompt learning for vision-language models

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Zi- wei Liu. Conditional prompt learning for vision-language models. In CVPR, pages 16816–16825, 2022. 2

    2022

  44. [52]

    Learning to prompt for vision-language models

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

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.