Pith. sign in

REVIEW 4 major objections 4 minor 74 references

BatStyler: Advancing Multi-category Style Generation for Source-free Domain Generalization

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read BatStyler claims that replacing fine-grained semantic constraints with coarse ones and seeding styles from a neural-collapse frame lifts source-free domain generalization on many-category benchmarks.

desk verdict Useful incremental SFDG method with real speedups and modest M-Avg gains, undermined by missing ViT-L/14 baselines and an overclaimed uniformity result. read the letter →

arxiv 2501.01109 v1 pith:XGI3C7T6 submitted 2025-01-02 cs.CV cs.AI

Xiusheng Xu , Lei Qi , Jingyang Zhou , Xin Geng This is my paper · ORCID
classification cs.CVcs.AI
keywords source-freedomaingeneralizationstylegenerationmulti-categoryclassificationvision-languagemodelsneuralcollapsepromptlearningCLIP
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 proposes BatStyler, a training method for source-free domain generalization that creates synthetic training features by optimizing pseudo-style word embeddings in CLIP's joint text-image space. It claims that on datasets with many categories, existing style synthesis collapses because enforcing semantic consistency across every category name compresses the space available for style variation, and because orthogonal style vectors cannot cover the embedding space. BatStyler replaces fine-grained category constraints with coarse-grained semantics extracted by clustering category names and querying a large language model, and seeds style vectors with a neural-collapse equiangular tight frame so they are uniformly spread and can be trained in parallel. The reported result is comparable accuracy on less-category benchmarks and higher average accuracy on multi-category benchmarks across three CLIP backbones, with style-generation training time cut to about a tenth of the earlier method on ImageNet-R.

What carries the argument

The load-bearing mechanism is the replacement of two constraints. The Coarse Semantic Generation module clusters class-name text features and extracts shared coarse labels, converting the sum over all N categories into a sum over a small set CSS, directly lowering the number of semantic-consistency constraints without weakening semantics into randomness. The Uniform Style Generation module defines K pseudo-style templates as the columns of an equiangular tight frame from neural collapse, satisfying $w_{n_1}^T w_{n_2} = \frac{K}{K-1}\delta_{n_1,n_2} - \frac{1}{K-1}$, packs them into a fixed classifier, and trains styles with cross-entropy so styles spread uniformly across the joint space and are trained in parallel.

What would settle it

Run the identical BatStyler pipeline on ImageNet-R but with CLIP's text encoder output projection randomly permuted before style generation, so the text and image spaces are no longer aligned; if accuracy stays near the reported 59.9 for ResNet-50, the joint-space transfer assumption is not load-bearing, whereas a large drop would confirm the paper's stated dependence on CLIP alignment.

Watch

Extended reading notes

Core claim

The paper's central claim is that the style-diversity bottleneck in source-free domain generalization is the semantic-consistency loss: as the number of classes N grows, enforcing each style-content prompt to align with every class name applies N parallel constraints that pull learned styles together, raising pairwise cosine similarity and shrinking diversity. BatStyler replaces those N fine-grained constraints with a small coarse-grained semantic set CSS obtained by clustering class-name text features with k-means++ and asking an LLM to describe each cluster, so semantic consistency is preserved with far fewer constraints. For coverage, it replaces orthogonality-based diversity with a fixed classifier whose K weight vectors are initialized as a neural-collapse equiangular tight frame, giving K equal-margin, uniformly distributed templates; because these templates form a fixed classifier, style training becomes a parallel cross-entropy problem rather than one-by-one orthogonalization. The paper reports that this combination yields the best multi-category average on ImageNet-R, DomainNet, and ImageNet-S for ResNet-50, ViT-B/16, and ViT-L/14 CLIP encoders, and that the first training stage takes roughly 10% of the baseline's time on ImageNet-R.

Load-bearing premise

The method assumes that CLIP's text and image encoders are aligned closely enough that a classifier trained on text features of synthetic style prompts transfers to real images simply by swapping the text encoder for the image encoder at inference; the authors state that if the two modalities are not well aligned, performance deteriorates.

Editorial extensions

If this is right

  • On multi-category benchmarks (ImageNet-R, DomainNet, ImageNet-S), BatStyler reports higher multi-category average accuracy than prior source-free methods on all three CLIP backbones tested.
  • On the less-category benchmarks (PACS, VLCS, OfficeHome) the method is comparable to, not ahead of, the strongest baselines, so the benefit is specific to the many-category regime.
  • The parallel style training makes the first training stage about 10% of the baseline's wall-clock time on ImageNet-R, at the cost of higher GPU memory.
  • Ablations attribute the larger share of the multi-category gain to the Coarse Semantic Generation module, implying that redundant fine-grained semantic constraints are the main obstacle.
  • The method remains fully source-free: only category names are used, no source-domain images are needed for training.

Reading between the lines

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

  • If the mechanism is correct, then any prompt-based data synthesis in CLIP's joint space should re-examine whether per-class consistency losses are silently limiting diversity; coarse-to-fine constraint reduction could transfer to other zero-shot and prompt-tuning settings.
  • The uniform-template idea suggests that the number of styles K can be pushed toward the embedding dimension without orthogonality collapse, making style count a tunable diversity knob rather than a fixed hyperparameter.
  • A testable extension is to vary the LLM query structure (hierarchical clusters, multiple granularities) to see whether the reported sweet spot of three coarse semantics per cluster generalizes to other datasets and backbones.
  • Because the paper's conclusion states the dependence on CLIP alignment, a sweep across other vision-language models with weaker text-image alignment would reveal whether the gains survive outside CLIP.
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 / 4 minor

Summary. The paper proposes BatStyler, a source-free domain generalization method built on CLIP that synthesizes pseudo-styles purely from text prompts. Two modules are introduced: Coarse Semantic Generation (CSG), which clusters category names and uses GPT-4 to extract coarse-grained semantics so that the semantic-consistency loss has fewer terms, and Uniform Style Generation (USG), which initializes a fixed classifier with neural-collapse ETF vectors to train pseudo-style word embeddings in parallel. The learned style-content text features are used to train a linear classifier that is transferred to the image encoder at inference. Experiments on PACS, VLCS, OfficeHome, ImageNet-R, DomainNet, and ImageNet-S report the best multi-category average accuracy on all three CLIP backbones (ResNet-50, ViT-B/16, ViT-L/14) and a large reduction in style-generation training time.

Significance. If the empirical results hold, BatStyler is a useful advance for SFDG in multi-category settings: it directly addresses the redundancy of fine-grained semantic constraints and parallelizes style training, with consistent gains over PromptStyler and other baselines on ImageNet-R, DomainNet, and ImageNet-S. The paper includes per-domain results, ablations, resource-usage tables, a sensitivity study for the number of pseudo-styles and coarse semantics, and t-SNE/text-to-image qualitative checks. The reliance on CLIP's joint vision-language space is explicitly acknowledged in Sec. V and is shared with all compared methods, so it is a limitation of the approach rather than a flaw in the comparison. However, the paper promises theoretical evidence that is not delivered, and the main comparison table has missing baselines and arithmetic inconsistencies that directly affect the headline M-Avg claim.

major comments (4)
  1. [Sec. IV-C, Table I (ViT-L/14 block)] The ViT-L/14 block of Table I omits PromptTA and DPStyler, the two strongest style-generation baselines in the same table for the smaller backbones. This conflicts with the statement in Sec. IV-C that 'In PromptStyler, PromptTA and DPStyler, we employ same configuration with BatStyler to conduct the comparison.' Because the central claim is that BatStyler surpasses state-of-the-art methods on multi-category datasets, the missing baselines leave the claim unverified on ViT-L/14: if either method improves over PromptStyler on ImageNet-R by a margin comparable to the ViT-B/16 row, its M-Avg could exceed BatStyler's 69.3. Please add the missing results or restrict the claim to the backbones where the comparison is complete.
  2. [Table I (Avg and M-Avg arithmetic)] Several reported Avg and M-Avg values are inconsistent with the per-dataset numbers. For example, PromptTA on ViT-B/16 has multi-category accuracies 75.8, 57.2, and 44.3, which average to 59.1, not the reported 59.9; DPStyler on ViT-B/16 averages to 60.0, not 59.8; DPStyler on ResNet-50 averages to 45.8, not 45.9. Since M-Avg is the headline metric for the paper's central claim, all such values must be recalculated and the conclusions rechecked.
  3. [Sec. III-B, Eq. (5)] The claim that the K=80 neural-collapse vectors are 'uniformly distributed throughout the entire joint space' is not correct as stated. An ETF of K vectors in R^P spans a subspace of dimension at most K (through the partial orthogonal matrix U in the ETF definition), so for P=1024 and K=80 the vectors leave most of the space uncovered, exactly as the paper notes for 80 orthogonal vectors. Please rephrase this to 'maximally separated within the subspace they span' or provide a formal sense in which the ETF covers the joint space more uniformly.
  4. [Sec. I and Sec. III-A] The introduction states that 'experimental results and theoretical evidence reveal that the semantic consistency compress the space of style diversity,' but no theoretical result is proved anywhere in the paper. The only support is the algebraic observation in Eq. (3) that R(θ) contains a sum over N categories, so replacing N with |css| reduces the number of terms; this is an algebraic observation, not theoretical evidence. Please either remove the phrase or supply an actual analysis (e.g., a bound on the achievable style diversity under the consistency constraints).
minor comments (4)
  1. [Abstract and Sec. I] The phrase 'an coarse semantic generation module' should be 'a coarse semantic generation module,' and 'Remakably' in Sec. IV-C should be 'Remarkably.'
  2. [Sec. III-B] 'schmidt orthogonalization' should be 'Gram-Schmidt orthogonalization.'
  3. [Abstract] The text says 'Code is available here' but no URL is provided; please include the actual link.
  4. [References and Table I] Reference [66] (DPStyler) is listed with Venue 'TMM'2024' in Table I, but the bibliography entry states it is an arXiv preprint; please align these entries.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: reported gains are measured on external image benchmarks, and internal style-diversity diagnostics are not the paper's headline prediction.

full rationale

BatStyler's central claim is accuracy on held-out image benchmarks (PACS, VLCS, OfficeHome, ImageNet-R, DomainNet, ImageNet-S), which is not fitted from or defined by the method's own parameters. The two proposed modules, CSG and USG, alter the style-generation loss, but the reported accuracies are obtained by training a linear classifier on synthetic text features and evaluating on real images, an external and independent measure. The internal diagnostic (average cosine similarity between synthetic styles) is closely related to the training objective, since the NC-initialized classifier and cross-entropy loss directly encourage uniform style directions, but this diagnostic is used as motivation and analysis, not as the paper's central evidence. The acknowledged limitation that performance depends on CLIP's vision-language alignment is a correctness risk shared by all compared methods and is explicitly stated, not a circular step. Self-citations, notably DPStyler (ref. 66) and NormAUG (ref. 59), appear only as comparison baselines and related work; they are not load-bearing for the derivation of BatStyler's losses or its accuracy claims. No equation in the paper reduces to the final reported result by construction, and no fitted parameter is renamed as a prediction. Therefore the derivation chain is self-contained with respect to the headline empirical claim.

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

The method introduces no new physical entities. Central claims rest on CLIP's joint-space alignment, the reliability of KMeans plus GPT-4 coarse semantics, the neural-collapse template as a style prior, and several hyperparameters (K=80, C=3, lambda=1). C is selected using test benchmark results, which is a selection-bias concern.

free parameters (4)
  • C, number of coarse semantics per cluster = 3
    Set after comparing C=1,3,5,10 on ImageNet-R, DomainNet, and ImageNet-S test sets (Table VI); C=10 degrades accuracy, so the value is chosen from test-set results.
  • K, number of pseudo-styles = 80
    Default set to match PromptStyler for comparison; Fig. 8 shows accuracy does not consistently improve with K, so this is a practical choice rather than a fitted optimum.
  • lambda, semantic consistency loss weight = 1
    Set to 1 after Fig. 7 shows performance is stable around 0.8; effectively L_SC has unit weight in Eq. 6.
  • KMeans cluster count = silhouette-optimal, dataset-dependent
    Chosen by silhouette coefficient on category text features; it determines the number of coarse semantics and hence the strength of semantic consistency, but is not tuned against test accuracy.
assumptions (4)
  • domain assumption CLIP text and image encoders share a joint space in which text-prompt features can serve as training data for image classification.
    Assumed throughout; Sec III-C replaces the text encoder with the image encoder at inference. The authors explicitly note in Sec V that poor vision-language alignment would hurt performance.
  • domain assumption The neural-collapse ETF (Eq. 5) is a beneficial initialization for pseudo-style vectors and produces broadly distributed styles.
    Used in Sec III-B; the paper provides no proof that cross-entropy to a fixed ETF classifier yields maximal style diversity, and the claim that 80 vectors cover a 1024-d joint space is not justified.
  • domain assumption Clustering category text features with KMeans++ and querying GPT-4 yields coarse semantics that preserve enough semantic consistency.
    Sec III-A relies on this for L_SC; no human evaluation or listing of extracted coarse semantics is provided.
  • ad hoc to paper The tradeoff in Eq. 1 between diversity loss and semantic consistency loss correctly models style generation, and reducing N is preferable to lowering lambda.
    Introduced in Sec III-A/D; motivated empirically by Table VII and Fig. 7, not derived from first principles.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BatStyler: Advancing Multi-category Style Generation for Source-free Domain Generalization." pith.science (2026). https://pith.science/paper/XGI3C7T6

@misc{pith2026250101109,
  author       = {Pith},
  title        = {Pith review of: BatStyler: Advancing Multi-category Style Generation for Source-free Domain Generalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XGI3C7T6}},
  note         = {Machine review of arXiv:2501.01109}
}
read the original abstract

Source-Free Domain Generalization (SFDG) aims to develop a model that performs on unseen domains without relying on any source domains. However, the implementation remains constrained due to the unavailability of training data. Research on SFDG focus on knowledge transfer of multi-modal models and style synthesis based on joint space of multiple modalities, thus eliminating the dependency on source domain images. However, existing works primarily work for multi-domain and less-category configuration, but performance on multi-domain and multi-category configuration is relatively poor. In addition, the efficiency of style synthesis also deteriorates in multi-category scenarios. How to efficiently synthesize sufficiently diverse data and apply it to multi-category configuration is a direction with greater practical value. In this paper, we propose a method called BatStyler, which is utilized to improve the capability of style synthesis in multi-category scenarios. BatStyler consists of two modules: Coarse Semantic Generation and Uniform Style Generation modules. The Coarse Semantic Generation module extracts coarse-grained semantics to prevent the compression of space for style diversity learning in multi-category configuration, while the Uniform Style Generation module provides a template of styles that are uniformly distributed in space and implements parallel training. Extensive experiments demonstrate that our method exhibits comparable performance on less-category datasets, while surpassing state-of-the-art methods on multi-category datasets.

Figures

Figures reproduced from arXiv: 2501.01109 by the authors.

Figure 1
Figure 1. Average cosine similarity (↓) of synthetic styles. Comparison of PromptStyler and BatStyler on three models: ResNet-50, ViT-B/16 and ViT￾L/14. We randomly sample 5, 200, 400, 600, 800 and 1000 category names from ImageNet-S. The number of style words is 80 and text features are obtained from prompt (e.g. “a S style of a”) through text encoder. 0 200 400 600 800 1000 0.0 0.5 1.0 1.5 2.0 Time (h) Ascending number of c… view at source ↗
Figure 2
Figure 2. Training time (↓) of first training stage. The experimental configuration adheres to the identical setup as described in [PITH_FULL_IMAGE:figures/full_fig_p001_2.png] view at source ↗
Figure 3
Figure 3. Overview of BatStyler. A Coarse Semantic Generation module is used to extract coarse-grained semantics of downstream categories. A classifier [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The overview of coarse semantic generation module (CSG). Here we [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: t-SNE [74] visualization result. The style-content features visualized by PromptStyler (Left) and BatStyler (Right) on four randomly selected categories from ImageNet-S. Different colors represent different categories. also be seen that the performance of BatStyler is …
Figure 6
Figure 6. Figure 6: Text-to-Image synthesis results using well-learned pseudo style words. We randomly select 6 style-content prompt, we set content as “dog”, and [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Performance on different [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Performance on different number of styles across ImageNet-R, [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 70 canonical work pages

  1. [1]

    Solid: minimizing tissue distortion for brain-wide profiling of diverse architec- tures,

    J. Zhu, X. Liu, Z. Liu, Y . Deng, J. Xu, K. Liu, R. Zhang, X. Meng, P. Fei, T. Yu et al. , “Solid: minimizing tissue distortion for brain-wide profiling of diverse architec- tures,” Nature Communications, p. 8303, 2024

    work page 2024

  2. [2]

    V oxelmorph: A learning framework for deformable medical image registration,

    G. Balakrishnan, A. Zhao, M. R. Sabuncu, J. V . Guttag, and A. V . Dalca, “V oxelmorph: A learning framework for deformable medical image registration,” IEEE Trans- actions on Medical Imaging , pp. 1788–1800, 2019

    work page 2019

  3. [3]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Conference on Com- puter Vision and Pattern Recognition (CVPR) , 2016, pp. 770–778

    work page 2016

  4. [4]

    Imagenet classification with deep convolutional neural networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional neural networks,” in Advances in Neural Information Processing Systems (NeurIPS), 2012, pp. 1106–1114

    work page 2012

  5. [5]

    Very deep convolu- tional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolu- tional networks for large-scale image recognition,” in International Conference on Learning Representations (ICLR), 2014

    work page 2014

  6. [6]

    Application of medical image detection tech- nology based on deep learning in pneumoconiosis diag- nosis,

    S. Peng, “Application of medical image detection tech- nology based on deep learning in pneumoconiosis diag- nosis,” Data Intelligence , vol. 5, pp. 1033–1047, 2023

    work page 2023

  7. [7]

    Densely connected convolutional networks,

    G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Wein- berger, “Densely connected convolutional networks,” in 11 Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 2261–2269

    work page 2017

  8. [8]

    Residual attention network for image classification,

    F. Wang, M. Jiang, C. Qian, S. Yang, C. Li, H. Zhang, X. Wang, and X. Tang, “Residual attention network for image classification,” in Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 6450–6458

    work page 2017

Show all 74 references
  1. [9]

    End-to-end object detection with transformers,

    N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kir- illov, and S. Zagoruyko, “End-to-end object detection with transformers,” in European Conference on Com- puter Vision (ECCV) , vol. 12346, 2020, pp. 213–229

    2020

  2. [10]

    Faster R-CNN: towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. B. Girshick, and J. Sun, “Faster R-CNN: towards real-time object detection with region proposal networks,” in Advances in Neural Information Processing Systems (NeurIPS) , 2015, pp. 91–99

    2015

  3. [11]

    One-shot adaptation of supervised deep convolutional models,

    J. Hoffman, E. Tzeng, J. Donahue, Y . Jia, K. Saenko, and T. Darrell, “One-shot adaptation of supervised deep convolutional models,” in International Conference on Learning Representations (ICLR) , 2014

    2014

  4. [12]

    Fifo: Learning fog- invariant features for foggy scene segmentation,

    S. Lee, T. Son, and S. Kwak, “Fifo: Learning fog- invariant features for foggy scene segmentation,” in Con- ference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 18 889–18 899

    2022

  5. [13]

    Semi-supervised domain adaptation via minimax en- tropy,

    K. Saito, D. Kim, S. Sclaroff, T. Darrell, and K. Saenko, “Semi-supervised domain adaptation via minimax en- tropy,” in International Conference on Computer Vision (ICCV), 2019, pp. 8049–8057

    2019

  6. [14]

    Domain generalization via encoding and resampling in a unified latent space,

    Y . Liu, Z. Xiong, Y . Li, X. Tian, and Z. Zha, “Domain generalization via encoding and resampling in a unified latent space,” IEEE Transactions on Multimedia (TMM) , vol. 25, pp. 126–139, 2023

    2023

  7. [15]

    Style nor- malization and restitution for domain generalization and adaptation,

    X. Jin, C. Lan, W. Zeng, and Z. Chen, “Style nor- malization and restitution for domain generalization and adaptation,” IEEE Transactions on Multimedia (TMM) , vol. 24, pp. 3636–3651, 2022

    2022

  8. [16]

    A novel mix- normalization method for generalizable multi-source per- son re-identification,

    L. Qi, L. Wang, Y . Shi, and X. Geng, “A novel mix- normalization method for generalizable multi-source per- son re-identification,” IEEE Transactions on Multimedia (TMM), vol. 25, pp. 4856–4867, 2023

    2023

  9. [17]

    Learning features of intra-consistency and inter-diversity: Keys toward gen- eralizable deepfake detection,

    H. Chen, Y . Lin, B. Li, and S. Tan, “Learning features of intra-consistency and inter-diversity: Keys toward gen- eralizable deepfake detection,” IEEE Transactions on Circuits and Systems for Video Technology (TCSVT) , vol. 33, pp. 1468–1480, 2023

    2023

  10. [18]

    Prompt- styler: Prompt-driven style generation for source-free domain generalization,

    J. Cho, G. Nam, S. Kim, H. Yang, and S. Kwak, “Prompt- styler: Prompt-driven style generation for source-free domain generalization,” in International Conference on Computer Vision (ICCV) , 2023, pp. 15 656–15 666

    2023

  11. [19]

    Towards data-free domain generalization,

    A. Frikha, H. Chen, D. Krompaß, T. A. Runkler, and V . Tresp, “Towards data-free domain generalization,” in ACML, 2022, pp. 327–342

    2022

  12. [20]

    Domain-unified prompt representations for source-free domain general- ization,

    H. Niu, H. Li, F. Zhao, and B. Li, “Domain-unified prompt representations for source-free domain general- ization,” arXiv preprint arXiv:2209.14926 , 2022

    2022 arXiv

  13. [21]

    Generalize then adapt: Source-free domain adaptive semantic segmentation,

    J. N. Kundu, A. R. Kulkarni, A. Singh, V . Jampani, and R. V . Babu, “Generalize then adapt: Source-free domain adaptive semantic segmentation,” in ICCV, 2021, pp. 7026–7036

    2021

  14. [22]

    Source- free unsupervised domain adaptation: A survey,

    Y . Fang, P. Yap, W. Lin, H. Zhu, and M. Liu, “Source- free unsupervised domain adaptation: A survey,” Neural Networks, vol. 174, p. 106230, 2024

    2024

  15. [23]

    Model adap- tation: Historical contrastive learning for unsupervised domain adaptation without source data,

    J. Huang, D. Guan, A. Xiao, and S. Lu, “Model adap- tation: Historical contrastive learning for unsupervised domain adaptation without source data,” in Advances in Neural Information Processing Systems (NeurIPS) , 2021, pp. 3635–3649

    2021

  16. [24]

    Source-free domain adaptation with frozen multimodal foundation model,

    S. Tang, W. Su, M. Ye, and X. Zhu, “Source-free domain adaptation with frozen multimodal foundation model,” in CVPR, 2024, pp. 23 711–23 720

    2024

  17. [25]

    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 International Conference on Machine Learning (ICML) , 2021, pp. 8748–8763

    2021

  18. [26]

    Adversarially adaptive normalization for single domain generalization,

    X. Fan, Q. Wang, J. Ke, F. Yang, B. Gong, and M. Zhou, “Adversarially adaptive normalization for single domain generalization,” in Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 8208–8217

    2021

  19. [27]

    Learning to learn single domain generalization,

    F. Qiao, L. Zhao, and X. Peng, “Learning to learn single domain generalization,” in Conference on Computer Vi- sion and Pattern Recognition (CVPR) , 2020, pp. 12 553– 12 562

    2020

  20. [28]

    Generalized semi-supervised and structured subspace learning for cross-modal retrieval,

    L. Zhang, B. Ma, G. Li, Q. Huang, and Q. Tian, “Generalized semi-supervised and structured subspace learning for cross-modal retrieval,” IEEE Transactions on Multimedia (TMM) , vol. 20, no. 1, pp. 128–141, 2018

    2018

  21. [29]

    Three heads better than one: Pure entity, relation label and adversarial training for cross-domain few-shot relation extraction,

    W. Fang, C. Ouyang, Q. Lin, and Y . Yuan, “Three heads better than one: Pure entity, relation label and adversarial training for cross-domain few-shot relation extraction,” Data Intelligence , vol. 5, pp. 807–823, 2023

    2023

  22. [30]

    Reduc- ing domain gap by reducing style bias,

    H. Nam, H. Lee, J. Park, W. Yoon, and D. Yoo, “Reduc- ing domain gap by reducing style bias,” in Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 8690–8699

    2021

  23. [31]

    Permuted adain: Reducing the bias towards global statistics in image classification,

    O. Nuriel, S. Benaim, and L. Wolf, “Permuted adain: Reducing the bias towards global statistics in image classification,” in Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 9482–9491

    2021

  24. [32]

    Domain generalization with optimal transport and metric learning,

    F. Zhou, Z. Jiang, C. Shui, B. Wang, and B. Chaib- draa, “Domain generalization with optimal transport and metric learning,” arXiv preprint arXiv:2007.10573, 2020

    2007 arXiv

  25. [33]

    Moment matching for multi-source domain adaptation,

    X. Peng, Q. Bai, X. Xia, Z. Huang, K. Saenko, and B. Wang, “Moment matching for multi-source domain adaptation,” in International Conference on Computer Vision (ICCV), 2019, pp. 1406–1415

    2019

  26. [34]

    3d-aided deep pose-invariant face recogni- tion,

    J. Zhao, L. Xiong, Y . Cheng, Y . Cheng, J. Li, L. Zhou, Y . Xu, J. Karlekar, S. Pranata, S. Shen, J. Xing, S. Yan, and J. Feng, “3d-aided deep pose-invariant face recogni- tion,” in IJCAI, 2018, pp. 1184–1190

    2018

  27. [35]

    Two at once: Enhancing learning and generalization capacities via ibn-net,

    X. Pan, P. Luo, J. Shi, and X. Tang, “Two at once: Enhancing learning and generalization capacities via ibn-net,” in European Conference on Computer Vision (ECCV), 2018, pp. 484–500

    2018

  28. [36]

    Frustratingly easy person re-identification: Generalizing person re-id in practice,

    J. Jia, Q. Ruan, and T. M. Hospedales, “Frustratingly easy person re-identification: Generalizing person re-id in practice,” in British Machine Vision Conference (BMVC), 2019, p. 117. 12

    2019

  29. [37]

    Cross-modal data augmentation for tasks of different modalities,

    D. Chen, Y . Zhuang, Z. Shen, C. Yang, G. Wang, S. Tang, and Y . Yang, “Cross-modal data augmentation for tasks of different modalities,” IEEE Transactions on Multimedia (TMM) , vol. 25, pp. 7814–7824, 2023

    2023

  30. [38]

    Dual-agent gans for photorealistic and identity preserving profile face synthesis,

    J. Zhao, L. Xiong, J. Karlekar, J. Li, F. Zhao, Z. Wang, S. Pranata, S. Shen, S. Yan, and J. Feng, “Dual-agent gans for photorealistic and identity preserving profile face synthesis,” in NeurIPS, 2017, pp. 66–76

    2017

  31. [39]

    Look across elapse: Disen- tangled representation learning and photorealistic cross- age face synthesis for age-invariant face recognition,

    J. Zhao, Y . Cheng, Y . Cheng, Y . Yang, F. Zhao, J. Li, H. Liu, S. Yan, and J. Feng, “Look across elapse: Disen- tangled representation learning and photorealistic cross- age face synthesis for age-invariant face recognition,” in AAAI, 2019, pp. 9251–9258

    2019

  32. [40]

    Scaling up visual and vision-language representation learning with noisy text supervision,

    C. Jia, Y . Yang, Y . Xia, Y . Chen, Z. Parekh, H. Pham, Q. V . Le, Y . Sung, Z. Li, and T. Duerig, “Scaling up visual and vision-language representation learning with noisy text supervision,” in ICML, 2021, pp. 4904–4916

    2021

  33. [41]

    BLIP: boot- strapping language-image pre-training for unified vision- language understanding and generation,

    J. Li, D. Li, C. Xiong, and S. C. H. Hoi, “BLIP: boot- strapping language-image pre-training for unified vision- language understanding and generation,” in International Conference on Machine Learning (ICML) , 2022, pp. 12 888–12 900

    2022

  34. [42]

    Minigpt-4: Enhancing vision-language understanding with advanced large language models,

    D. Zhu, J. Chen, X. Shen, X. Li, and M. Elhoseiny, “Minigpt-4: Enhancing vision-language understanding with advanced large language models,” arXiv preprint arXiv:2304.10592, 2023

    2023 arXiv

  35. [43]

    Faster zero-shot multi-modal entity linking via visual-linguistic representation,

    Q. Zheng, H. Wen, M. Wang, G. Qi, and C. Bai, “Faster zero-shot multi-modal entity linking via visual-linguistic representation,” Data Intelligence , vol. 4, pp. 493–508, 2022

    2022

  36. [44]

    Diagnosing and rectifying vision models using language,

    Y . Zhang, J. Z. HaoChen, S. Huang, K. Wang, J. Zou, and S. Yeung, “Diagnosing and rectifying vision models using language,” in International Conference on Learn- ing Representations (ICLR) , 2023

    2023

  37. [45]

    Prevalence of neural collapse during the terminal phase of deep learning training,

    V . Papyan, X. Han, and D. L. Donoho, “Prevalence of neural collapse during the terminal phase of deep learning training,” Proceedings of the National Academy of Sciences (PNAS) , 2020

    2020

  38. [46]

    Inducing neural collapse in imbalanced learning: Do we really need a learnable classifier at the end of deep neural network?

    Y . Yang, S. Chen, X. Li, L. Xie, Z. Lin, and D. Tao, “Inducing neural collapse in imbalanced learning: Do we really need a learnable classifier at the end of deep neural network?” in Advances in Neural Information Processing Systems (NeurIPS) , 2022

    2022

  39. [47]

    Neural collapse in- spired attraction-repulsion-balanced loss for imbalanced learning,

    L. Xie, Y . Yang, D. Cai, and X. He, “Neural collapse in- spired attraction-repulsion-balanced loss for imbalanced learning,” Neurocomputing, vol. 527, pp. 60–70, 2023

    2023

  40. [48]

    Targeted representation alignment for open-world semi-supervised learning,

    R. Xiao, L. Feng, K. Tang, J. Zhao, Y . Li, G. Chen, and H. Wang, “Targeted representation alignment for open-world semi-supervised learning,” in CVPR, 2024, pp. 23 072–23 082

    2024

  41. [49]

    Neural collapse inspired semi-supervised learning with fixed classifier,

    Z. Hu, Y . Wang, H. Ning, Y . Tai, and F. Nie, “Neural collapse inspired semi-supervised learning with fixed classifier,” Information Sciences , vol. 667, p. 120469, 2024

    2024

  42. [50]

    Learning optimal inter-class margin adaptively for few- shot class-incremental learning via neural collapse-based meta-learning,

    H. Ran, W. Li, L. Li, S. Tian, X. Ning, and P. Tiwari, “Learning optimal inter-class margin adaptively for few- shot class-incremental learning via neural collapse-based meta-learning,” Information Processing and Manage- ment, p. 103664, 2024

    2024

  43. [51]

    Neural collapse inspired feature-classifier alignment for few-shot class-incremental learning,

    Y . Yang, H. Yuan, X. Li, Z. Lin, P. H. S. Torr, and D. Tao, “Neural collapse inspired feature-classifier alignment for few-shot class-incremental learning,” in International Conference on Learning Representations (ICLR) , 2023

    2023

  44. [52]

    Neural collapse anchored prompt tuning for generalizable vision-language models,

    D. Zhu, Z. Li, M. Zhang, J. Yuan, J. Liu, K. Kuang, and C. Wu, “Neural collapse anchored prompt tuning for generalizable vision-language models,” in KDD, 2024, pp. 4631–4640

    2024

  45. [53]

    Bridging the gap: neural collapse inspired prompt tuning for generalization under class imbalance,

    D. Zhu, Y . Li, M. Zhang, J. Yuan, J. Liu, K. Kuang, and C. Wu, “Bridging the gap: neural collapse inspired prompt tuning for generalization under class imbalance,” arXiv preprint arXiv:2306.15955 , 2023

    2023 arXiv

  46. [54]

    k-means++: the advan- tages of careful seeding,

    D. Arthur and S. Vassilvitskii, “k-means++: the advan- tages of careful seeding,” in Symposium on Discrete Algorithms (ACM-SIAM), 2007, pp. 1027–1035

    2007

  47. [55]

    Gpt-4 technical report,

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat et al., “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774, 2023

    2023 arXiv

  48. [56]

    Arcface: Ad- ditive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Ad- ditive angular margin loss for deep face recognition,” in Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 4690–4699

    2019

  49. [57]

    Learning class and domain augmen- tations for single-source open-domain generalization,

    P. Bele, V . Bundele, A. Bhattacharya, A. Jha, G. Roig, and B. Banerjee, “Learning class and domain augmen- tations for single-source open-domain generalization,” in WACV, 2024, pp. 1805–1815

    2024

  50. [58]

    Exploring ex- plicitly disentangled features for domain generalization,

    J. Li, Y . Li, H. Wang, C. Liu, and J. Tan, “Exploring ex- plicitly disentangled features for domain generalization,” IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), vol. 33, pp. 6360–6373, 2023

    2023

  51. [59]

    Normaug: Normalization-guided augmentation for domain general- ization,

    L. Qi, H. Yang, Y . Shi, and X. Geng, “Normaug: Normalization-guided augmentation for domain general- ization,” IEEE Transactions on Image Processing (TIP) , pp. 1419–1431, 2024

    2024

  52. [60]

    Instance paradigm contrastive learning for domain gen- eralization,

    Z. Chen, W. Wang, Z. Zhao, F. Su, A. Men, and Y . Dong, “Instance paradigm contrastive learning for domain gen- eralization,” IEEE Transactions on Circuits and Systems for Video Technology (TCSVT) , vol. 34, pp. 1032–1042, 2024

    2024

  53. [61]

    Source-free domain adaptation with unrestricted source hypothesis,

    J. He, L. Wu, C. Tao, and F. Lv, “Source-free domain adaptation with unrestricted source hypothesis,” Pattern Recognit., vol. 149, p. 110246, 2024

    2024

  54. [62]

    Neighborhood-aware mutual information maximization for source-free domain adaptation,

    L. Zhang, Y . Wang, R. Song, M. Zhang, X. Li, and W. Zhang, “Neighborhood-aware mutual information maximization for source-free domain adaptation,” IEEE Transactions on Multimedia (TMM) , vol. 26, pp. 9564– 9574, 2024

    2024

  55. [63]

    Waffling around for performance: Visual classification with random words and broad concepts,

    K. Roth, J. Kim, A. S. Koepke, O. Vinyals, C. Schmid, and Z. Akata, “Waffling around for performance: Visual classification with random words and broad concepts,” in International Conference on Computer Vision (ICCV) , 2023, pp. 15 700–15 711

    2023

  56. [64]

    Stylip: Multi-scale style-conditioned prompt learning for clip-based domain generalization,

    S. Bose, A. Jha, E. Fini, M. Singha, E. Ricci, and B. Banerjee, “Stylip: Multi-scale style-conditioned prompt learning for clip-based domain generalization,” in WACV, 2024, pp. 5530–5540. 13

    2024

  57. [65]

    Promptta: Prompt-driven text adapter for source-free do- main generalization,

    H. Zhang, S. Bai, W. Zhou, J. Fu, and B. Chen, “Promptta: Prompt-driven text adapter for source-free do- main generalization,” arXiv preprint arXiv:2409.14163 , 2024

    2024 arXiv

  58. [66]

    Dpstyler: Dy- namic promptstyler for source-free domain generaliza- tion,

    Y . Tang, Y . Wan, L. Qi, and X. Geng, “Dpstyler: Dy- namic promptstyler for source-free domain generaliza- tion,” arXiv preprint arXiv:2403.16697 , 2024

    2024 arXiv

  59. [67]

    Visual classification via description from large language models,

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

    2023

  60. [68]

    Chatgpt-powered hierarchi- cal comparisons for image classification,

    Z. Ren, Y . Su, and X. Liu, “Chatgpt-powered hierarchi- cal comparisons for image classification,” in NeurIPS, A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, Eds., 2023

    2023

  61. [69]

    Deeper, broader and artier domain generalization,

    D. Li, Y . Yang, Y . Song, and T. M. Hospedales, “Deeper, broader and artier domain generalization,” in Interna- tional Conference on Computer Vision (ICCV) , 2017, pp. 5543–5551

    2017

  62. [70]

    Unbiased metric learning: On the utilization of multiple datasets and web images for softening bias,

    C. Fang, Y . Xu, and D. N. Rockmore, “Unbiased metric learning: On the utilization of multiple datasets and web images for softening bias,” in International Conference on Computer Vision (ICCV) , 2013, pp. 1657–1664

    2013

  63. [71]

    Deep hashing network for unsupervised do- main adaptation,

    H. Venkateswara, J. Eusebio, S. Chakraborty, and S. Pan- chanathan, “Deep hashing network for unsupervised do- main adaptation,” in Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 5385–5394

    2017

  64. [72]

    The many faces of robustness: A critical analysis of out-of-distribution gen- eralization,

    D. Hendrycks, S. Basart, N. Mu, S. Kadavath, F. Wang, E. Dorundo, R. Desai, T. Zhu, S. Parajuli, M. Guo, D. Song, J. Steinhardt, and J. Gilmer, “The many faces of robustness: A critical analysis of out-of-distribution gen- eralization,” in International Conference on Computer ...

    2021

  65. [73]

    Large-scale unsupervised semantic segmentation,

    S. Gao, Z. Li, M. Yang, M. Cheng, J. Han, and P. H. S. Torr, “Large-scale unsupervised semantic segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence Information (TPAMI) , vol. 45, no. 6, pp. 7457–7476, 2023

    2023

  66. [74]

    Visualizing data using t-sne,

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne,” Journal of Machine Learning Research (JMLR) , vol. 9, no. 11, 2008

    2008

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.