Pith. sign in

REVIEW 3 major objections 4 minor 60 references

Dual-Path Stable Soft Prompt Generation for Domain Generalization

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Negative prompts enlarge the decision margin and stabilize prompt-based domain generalization.

desk verdict A solid empirical recipe for stabilizing dynamic prompts via negative learning, but the margin theory rests on an unverified inequality the paper never measures. read the letter →

arxiv 2505.18770 v1 pith:2L6G6K2R submitted 2025-05-24 cs.CV cs.LG

classification cs.CVcs.LG
keywords domaingeneralizationvision-languagemodelspromptlearningsoftgenerationnegativevariabilitytransformergeneratorCLIP
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 targets a problem it calls Prompt Variability: dynamic prompt generators for vision-language models often produce very different prompts for the same image across random seeds, and those prompts sit far from the optimal domain prompt. The proposed method, Dual-Path Stable Soft Prompt Generation (DPSPG), trains two transformer-based prompt generators, one producing positive prompts such as “a photo of a {class}” and one producing negative prompts such as “a photo without a {class}”, then subtracts the negative score from the positive score at inference. The paper argues that negative learning enlarges the effective classification margin by at least $\alpha\delta$ and exponentially tightens an upper bound on the gradient norm, which explains smoother optimization and more robust prompts. Experiments on five domain-generalization benchmarks report higher average accuracy than prior prompt-generation methods together with lower seed-to-seed variance, supporting the claim that negative prompts are a useful stabilizer.

What carries the argument

The central object is the dual-path prompt generator: two transformer encoders, $G^+$ and $G^-$, each composed of four transformer layers and a linear layer, which map CLIP image embeddings to positive and negative soft prompts. These generators are trained with a mean-squared-error loss to match per-domain positive and negative prompt labels learned in a first stage, and at inference the two prompts define the combined logit $g_i(x)=s^+_i(x)-\alpha s^-_i(x)$. The load-bearing identity is the separation inequality $s^-_i(x)\ge s^-_y(x)+\delta$ for every wrong class $i$; substituting it into the margin equation yields $\Delta_i(x)\ge \Delta^+_i(x)+\alpha\delta$, which in turn produces the exponential Jacobian bound $\|J_f(x)\le \frac{L}{\tau}e^{-(\Delta^+_i(x)+\alpha\delta)/\tau}$ that carries the stability argument.

What would settle it

On any held-out target domain, compute $\delta(x)=\min_{i\ne y}\left(s^-_i(x)-s^-_y(x)\right)$ using the trained negative prompt generator. If a substantial fraction of target images have $\delta(x)\le 0$, then the inequality underlying Equation (12) fails on the test distribution, and the claimed margin expansion and exponential gradient-norm bound cannot explain the reported accuracy gains.

Watch

Extended reading notes

Core claim

The central claim is that negative learning stabilizes prompt generation. By generating a negative soft prompt alongside the positive one and using the combined logit $g_i(x)=s^+_i(x)-\alpha s^-_i(x)$, the margin between the true class and each incorrect class becomes at least $\Delta^+_i(x)+\alpha\delta$ whenever the negative-prompt separation condition $s^-_i(x)\ge s^-_y(x)+\delta$ holds for every incorrect class $i$. The paper further derives an exponential gradient-norm bound $\|J_f(x)\le \frac{L}{\tau}e^{-(\Delta^+_i(x)+\alpha\delta)/\tau}$, arguing that the enlarged margin yields smoother optimization and reduced sensitivity to input perturbations. Empirically, DPSPG reports state-of-the-art accuracy on PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet while producing prompts that cluster tightly around the optimal prompt, directly addressing the Prompt Variability identified in prior dynamic prompt methods.

Load-bearing premise

The margin and gradient-stability theorems rely on the assumption that for every wrong class, the negative prompt gives the true class a score at least $\delta$ lower than the wrong class, but this separation is learned only on source domains and is never verified on the unseen target domains where it must hold.

Editorial extensions

If this is right

  • If the margin result holds, dynamic prompt generation for domain generalization becomes reproducible: the same image will produce nearly identical prompts across seeds, removing seed dependence as a major source of performance variance.
  • The exponential gradient-norm bound implies smoother optimization, so DPSPG should train more stably and with less sensitivity to learning-rate and warm-up choices than GAN-based prompt generators such as SPG.
  • The dual-path recipe of training a positive and a negative generator on domain prompt labels and subtracting scores at inference is a general template that can be applied to other CLIP-based classification settings beyond the five DG benchmarks tested.
  • Because the margin gain scales with $\alpha\delta$, the balancing weight $\alpha$ acts as a confidence lever: larger values widen margins when negative prompts are reliable but increase reliance on source-learned negative labels.

Reading between the lines

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

  • The paper never measures $\delta$ on target data; a direct test is to compute $s^-_i(x)-s^-_y(x)$ for the learned negative prompts on held-out domains. If $\delta\le 0$ for a non-negligible fraction of test images, the claimed margin expansion is a source-domain artifact and the accuracy gains would need another explanation.
  • The dual-path stabilization mechanism is not specific to prompts: any generative adapter that conditions on image features could in principle gain seed stability from a secondary generator trained to reject the class-conditional template, although the paper does not explore that extension.
  • The robustness claim that a smaller Jacobian norm reduces sensitivity to input perturbations suggests DPSPG should exhibit improved adversarial robustness, but the paper runs no adversarial attacks; an adversarial-robustness evaluation would be a direct test of that corollary.
  • Fixing $\alpha=0.2$ across all datasets may leave performance on the table for large-shift domains such as Sketch and TerraIncognita, where the optimal balance between positive and negative scores could differ.
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

3 major / 4 minor

Summary. The paper proposes Dual-Path Stable Soft Prompt Generation (DPSPG), a two-stage method for domain generalization with CLIP-based prompt learning. In the first stage, domain-specific positive and negative soft-prompt labels are learned with cross-entropy and binary cross-entropy losses. In the second stage, two transformer-based generators produce instance-specific positive and negative prompts, trained by MSE alignment to those labels. At inference, the logit for class i is the positive-prompt similarity minus α times the negative-prompt similarity. The authors argue that this dual-path design reduces 'Prompt Variability' across random seeds, and they provide a theoretical analysis claiming that negative prompts enlarge the decision margin by at least αδ and tighten a gradient-norm bound. Experiments on PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet with ResNet-50 and ViT-B/16 backbones show consistent average accuracy improvements over prior prompt-based DG methods, along with lower standard deviations on several stability metrics. The code is publicly available.

Significance. If the empirical claims hold, DPSPG is a useful and well-engineered extension of dynamic prompt learning for DG: it introduces negative prompts inside a transformer-based generator, addresses a real phenomenon (seed-dependent prompt variability), and validates the design with ablations and stability analyses. The paper is commendable for releasing code, evaluating on five standard benchmarks with two backbones, and reporting training-time efficiency and stability comparisons. However, the theoretical centerpiece is not established: the margin-enlargement result is conditional on an asserted inequality that no training objective enforces and no experiment measures, and the 'gradient norm stability' bound concerns the input-output Jacobian rather than the parameter gradients actually optimized during training. The empirical gains, while consistently positive on average, are often small relative to the prior SPG baseline and are reported without per-run variance in the main tables. The conceptual contribution of 'Prompt Variability' is interesting, but the theoretical framing as currently written overstates what has been proven.

major comments (3)
  1. [§III-C, Eq. (11)] The entire margin-enlargement result (Eq. 12) and the subsequent exponential gradient bound (Eq. 17) rest on the inequality s_i^-(x) ≥ s_y^-(x) + δ for every incorrect class i and every input x. This inequality is introduced with the phrase 'by designing the negative prompts,' but none of the training objectives in §III-B enforces it: Eq. (5) is a BCE loss on label probabilities, and Eq. (6) is an MSE loss on prompt vectors. Neither is a ranking or margin loss defined on the inference-time negative scores. The paper presents no measurement of δ on source or target data, and no ablation checks whether the inequality holds for the generated negative prompts. If Eq. (11) fails for even one incorrect class, the direction of Eq. (12) reverses and negative learning can shrink the margin. This is a load-bearing assumption, not a derived result; please either enforce Eq. (11) with a suitable loss, empirically verify δ on both source and target splits, or explicitly reframe the theory as conditional on an assumption that is tested rather than asserted.
  2. [§III-C, Eqs. (13)–(17)] The claim of 'gradient norm stability' during optimization does not follow from the presented calculation. Eq. (13)–(16) bound the Jacobian of the softmax output with respect to the input x, ∂f/∂x, after assuming an L-Lipschitz map from x to the logits g. But the training in §III-B optimizes the parameters of G+ and G−; the quantity relevant for optimization is ∂L/∂θ, which is not controlled by the input-output Jacobian bound. A smaller ∥J_f(x)∥ is a robustness statement about sensitivity to input perturbations, not a statement about smoother or more stable gradient descent over prompt-generator parameters. Please correct the object being bounded, or remove the claim that this result explains stable optimization.
  3. [Tables II–IV] The main results are reported as single numbers without standard deviations over random seeds, despite the paper's emphasis on seed variability. Some of the headline gains are small relative to the closest baseline (e.g., Tables II–IV: VLCS ResNet-50 improvement over SPG is 0.13, PACS ViT-B/16 improvement is 0.25 and ties with VPT at 96.93, DomainNet ResNet-50 improvement is 0.26), and several per-domain accuracies are actually lower than SPG (e.g., VLCS Pascal 84.80 vs. 85.37; VLCS Sun 80.41 vs. 82.40; OfficeHome Product 82.60 vs. 83.40 for ResNet-50). Without variance estimates or significance tests, the claim of 'consistently outperforms' is not fully supported in the main tables. Please add mean±std across at least three seeds for the main comparisons and discuss the significance of the average gains.
minor comments (4)
  1. [§III-C, Eq. (5)] The description of y_dj as a 'multi-label one-hot encoded vector (e.g., [1,0,1])' is confusing because all DG datasets used in the paper are single-label; please clarify whether this is a one-hot vector or a true multi-label vector, and how the BCE loss is applied in the single-label case.
  2. [Table V] The row descriptions for the ablation are ambiguous. Row #3 is described in the text as 'DPSPG without incorporating negative learning,' but the checkmark pattern appears to include CGAN and exclude the transformer, which would make it an SPG variant rather than DPSPG without negative learning. Please clarify the exact component configuration of each row, for example with explicit '✓/✗' entries for all four components.
  3. [§III-B, Eqs. (6) and (7)] The symbol α is used both as the weight for the negative MSE loss in Eq. (6) and as the weight of the negative score in the inference logit in Eq. (7). The implementation section states α=0.2, but the sensitivity analysis (Fig. 6) is described as varying 'the combination weight α used in Equation 6.' Please clarify whether the same α is used in both places and whether the sensitivity analysis varies both simultaneously.
  4. [Abstract and Introduction] The abstract and introduction state that 'theoretical and empirical analyses demonstrate' that negative learning increases the effective margin and reduces the gradient-norm upper bound. Given the conditional nature of the theory (see major comment 1), please temper this phrasing, e.g., by saying the analysis shows the margin gain under an explicit separation condition that is then validated empirically.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is conditional and the empirical claims are validated on independent benchmarks.

full rationale

The paper's central theoretical claim (margin enlargement by αδ, Eq. 12, and the gradient-norm bound, Eq. 17) is a conditional derivation: substituting assumption (11), s−_i(x) ≥ s−_y(x) + δ, into Eq. (10) gives the margin bound by algebra. This is a legitimate conditional implication, not a circular reduction, because Eq. (11) is an explicit premise ('by designing the negative prompts') rather than the conclusion being reused as an input. The unverified status of Eq. (11) on target domains is a rigor or assumption gap, not circularity; the paper never measures δ and no training loss directly enforces it, but this does not make the derivation equivalent to its inputs. The empirical evaluation is conducted on five external domain-generalization benchmarks against published baselines, including SPG (ECCV 2024), which is a prior work with overlapping authors but is used as a comparison method, not as justification for the paper's theoretical or empirical claims. Self-citations in related work (e.g., Refs. 10, 14, 17, 34) are contextual and not load-bearing. No fitted parameter is renamed as a prediction, and the framework is not a re-labeling of a known result. Therefore the derivation chain is self-contained, with any concerns about Eq. (11) belonging to correctness risk rather than circularity.

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

The main free choices are alpha (0.2), prompt context length (4), and the unmeasured separation delta in the theoretical bound. The framework assumes source-domain prompt labels transfer to unseen domains and that the logit map is Lipschitz. No new physical or natural entities are postulated; the negative generator is an architectural component.

free parameters (3)
  • alpha = 0.2
    Balancing weight for negative prompts in training and inference. Set to 0.2 after a sensitivity sweep (Figure 6); the claimed margin gain is alpha times delta, so this parameter directly controls the theoretical benefit.
  • delta
    Hypothetical separation lower bound in Eq. (11). The paper never measures delta on target domains, so the quantitative margin claim depends on an unknown constant.
  • Context length L of soft prompts = 4
    Template context length fixed to 4 in the implementation details; affects the capacity of the learned prompts and is a hand-chosen hyperparameter.
assumptions (4)
  • ad hoc to paper Negative prompts satisfy s_i^-(x) >= s_y^-(x) + delta for every incorrect class i and every input x (Eq. 11).
    The margin lower bound in Eq. (12) and the gradient bound in Eq. (17) both follow from this inequality. It is asserted as a design property rather than proved for unseen target domains.
  • domain assumption The model's input-to-logit map is L-Lipschitz continuous.
    Used in Eq. (16) to convert logit sensitivity into input sensitivity. Standard but stated without a specific Lipschitz constant or proof.
  • domain assumption Source-domain positive and negative prompt labels transfer to unseen domains, so regressing generators to them improves target-domain accuracy.
    This is the core transfer assumption of the two-stage training. It is validated only indirectly through benchmark accuracy.
  • standard math Standard softmax/CLIP cosine-similarity formulation as in Eq. (2).
    Foundation of the margin and gradient analysis; taken from CLIP and the prompt-tuning literature.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dual-Path Stable Soft Prompt Generation for Domain Generalization." pith.science (2026). https://pith.science/paper/2L6G6K2R

@misc{pith2026250518770,
  author       = {Pith},
  title        = {Pith review of: Dual-Path Stable Soft Prompt Generation for Domain Generalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2L6G6K2R}},
  note         = {Machine review of arXiv:2505.18770}
}
read the original abstract

Domain generalization (DG) aims to learn a model using data from one or multiple related but distinct source domains that can generalize well to unseen out-of-distribution target domains. Inspired by the success of large pre-trained vision-language models (VLMs), prompt tuning has emerged as an effective generalization strategy. However, it often struggles to capture domain-specific features due to its reliance on manually or fixed prompt inputs. Recently, some prompt generation methods have addressed this limitation by dynamically generating instance-specific and domain-specific prompts for each input, enriching domain information and demonstrating potential for enhanced generalization. Through further investigation, we identify a notable issue in existing prompt generation methods: the same input often yields significantly different and suboptimal prompts across different random seeds, a phenomenon we term Prompt Variability. To address this, we introduce negative learning into the prompt generation process and propose Dual-Path Stable Soft Prompt Generation (DPSPG), a transformer-based framework designed to improve both the stability and generalization of prompts. Specifically, DPSPG incorporates a complementary prompt generator to produce negative prompts, thereby reducing the risk of introducing misleading information. Both theoretical and empirical analyses demonstrate that negative learning leads to more robust and effective prompts by increasing the effective margin and reducing the upper bound of the gradient norm. Extensive experiments on five DG benchmark datasets show that DPSPG consistently outperforms state-of-the-art methods while maintaining prompt stability.

Figures

Figures reproduced from arXiv: 2505.18770 by the authors.

Figure 1
Figure 1. Comparison of the inference stage between our proposed DPSPG [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison of prompt generation quality between our proposed [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The training process of DPSPG consists of two stages. In the first stage, positive and negative domain prompt labels are learned. In the second stage, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Two examples during inference. Compared with SPG, DPSPG [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Standard deviation of leave-one-domain-out accuracies across five datasets for various CLIP-based prompt learning methods using (a) ResNet-50 and [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Sensitivity analysis of parameter α on five DG benchmark datasets for multi-source DG performance with ResNet50 as the backbone. (#2 vs. #4) leads to a drop in accuracy. The transformer back￾bone better captures long-range dependencies and domain￾specific nuances, resu…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 36 canonical work pages

  1. [1]

    Domain general- ization: A survey,

    K. Zhou, Z. Liu, Y . Qiao, T. Xiang, and C. C. Loy, “Domain general- ization: A survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 4, pp. 4396–4415, 2022

  2. [2]

    Generalizing to unseen domains: A survey on domain generalization,

    J. Wang, C. Lan, C. Liu, Y . Ouyang, T. Qin, W. Lu, Y . Chen, W. Zeng, and S. Y . Philip, “Generalizing to unseen domains: A survey on domain generalization,” IEEE transactions on knowledge and data engineering , vol. 35, no. 8, pp. 8052–8072, 2022

  3. [3]

    Generalizing to unseen domains via adversarial data augmentation,

    R. V olpi, H. Namkoong, O. Sener, J. C. Duchi, V . Murino, and S. Savarese, “Generalizing to unseen domains via adversarial data augmentation,” Advances in neural information processing systems , vol. 31, 2018

  4. [4]

    A simple feature augmentation for domain generalization,

    P. Li, D. Li, W. Li, S. Gong, Y . Fu, and T. M. Hospedales, “A simple feature augmentation for domain generalization,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 8886–8895

  5. [5]

    mixup: Beyond empirical risk minimization,

    H. Zhang, M. Cisse, Y . N. Dauphin, and D. Lopez-Paz, “mixup: Beyond empirical risk minimization,” 2018. [Online]. Available: https://arxiv.org/abs/1710.09412

  6. [6]

    Domain generalization via invariant feature representation,

    K. Muandet, D. Balduzzi, and B. Sch ¨olkopf, “Domain generalization via invariant feature representation,” in International conference on machine learning. PMLR, 2013, pp. 10–18

  7. [7]

    Domain generalization with small data,

    K. Chen, E. Gal, H. Yan, and H. Li, “Domain generalization with small data,” International Journal of Computer Vision , vol. 132, no. 8, pp. 3172–3190, 2024

  8. [8]

    Ensemble of averages: Improving model selection and boosting performance in domain gener- alization,

    D. Arpit, H. Wang, Y . Zhou, and C. Xiong, “Ensemble of averages: Improving model selection and boosting performance in domain gener- alization,” Advances in Neural Information Processing Systems , vol. 35, pp. 8265–8277, 2022

Show all 60 references
  1. [9]

    Domain adaptation via prompt learning,

    C. Ge, R. Huang, M. Xie, Z. Lai, S. Song, S. Li, and G. Huang, “Domain adaptation via prompt learning,” IEEE Transactions on Neural Networks and Learning Systems , 2023

  2. [10]

    Prompt-based distribution alignment for unsupervised domain adaptation,

    S. Bai, M. Zhang, W. Zhou, S. Huang, Z. Luan, D. Wang, and B. Chen, “Prompt-based distribution alignment for unsupervised domain adaptation,” in Proceedings of the 38th AAAI Conference on Artificial Intelligence (AAAI 2024). AAAI Press , 2024

  3. [11]

    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 . PMLR, 2021, pp. 8748–8763

  4. [12]

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

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

  5. [13]

    Learning to prompt for vision- language models,

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

  6. [14]

    Revisiting the adversarial robustness of vision language models: a multimodal perspective,

    W. Zhou, S. Bai, D. P. Mandic, Q. Zhao, and B. Chen, “Revisiting the adversarial robustness of vision language models: a multimodal perspective,” arXiv preprint arXiv:2404.19287 , 2024

  7. [15]

    Maple: Multi-modal prompt learning,

    M. U. Khattak, H. Rasheed, M. Maaz, S. Khan, and F. S. Khan, “Maple: Multi-modal prompt learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 19 113–19 122

  8. [16]

    Domain prompt learning for efficiently adapting clip to unseen domains,

    X. Zhang, S. S. Gu, Y . Matsuo, and Y . Iwasawa, “Domain prompt learning for efficiently adapting clip to unseen domains,” Transactions of the Japanese Society for Artificial Intelligence , vol. 38, no. 6, pp. B–MC2 1, 2023

  9. [17]

    Soft prompt generation for domain generalization,

    S. Bai, Y . Zhang, W. Zhou, Z. Luan, and B. Chen, “Soft prompt generation for domain generalization,” in European Conference on Computer Vision, 2024

  10. [18]

    Cbda: Contrastive-based data augmentation for domain generalization,

    Z. Jiang, L. Zhang, X. Liang, and Z. Chen, “Cbda: Contrastive-based data augmentation for domain generalization,” IEEE Transactions on Computational Social Systems , 2024

  11. [19]

    Mixup-induced domain extrapolation for domain generalization,

    M. Cao and S. Chen, “Mixup-induced domain extrapolation for domain generalization,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 10, 2024, pp. 11 168–11 176

  12. [20]

    Domain generalization with adversarial feature learning,

    H. Li, S. J. Pan, S. Wang, and A. C. Kot, “Domain generalization with adversarial feature learning,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 5400–5409

  13. [21]

    Domain generalization via inter- domain alignment and intra-domain expansion,

    J. Hu, L. Qi, J. Zhang, and Y . Shi, “Domain generalization via inter- domain alignment and intra-domain expansion,” Pattern Recognition , vol. 146, p. 110029, 2024

  14. [22]

    Domain-adversarial training of neural networks,

    Y . Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Lavi- olette, M. March, and V . Lempitsky, “Domain-adversarial training of neural networks,” Journal of machine learning research, vol. 17, no. 59, pp. 1–35, 2016

  15. [23]

    Deep domain generalization via conditional invariant adversarial networks,

    Y . Li, X. Tian, M. Gong, Y . Liu, T. Liu, K. Zhang, and D. Tao, “Deep domain generalization via conditional invariant adversarial networks,” in Proceedings of the European conference on computer vision (ECCV) , 2018, pp. 624–639

  16. [24]

    Invariant risk minimization,

    M. Arjovsky, L. Bottou, I. Gulrajani, and D. Lopez-Paz, “Invariant risk minimization,” arXiv preprint arXiv:1907.02893 , 2019

  17. [25]

    Invariant information bottleneck for domain generalization,

    B. Li, Y . Shen, Y . Wang, W. Zhu, D. Li, K. Keutzer, and H. Zhao, “Invariant information bottleneck for domain generalization,” in Pro- ceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 7, 2022, pp. 7399–7407

  18. [26]

    Exploiting domain- specific features to enhance domain generalization,

    M.-H. Bui, T. Tran, A. Tran, and D. Phung, “Exploiting domain- specific features to enhance domain generalization,” Advances in Neural Information Processing Systems , vol. 34, pp. 21 189–21 201, 2021

  19. [27]

    Simple: Specialized model-sample matching for domain generalization,

    Z. Li, K. Ren, X. Jiang, Y . Shen, H. Zhang, and D. Li, “Simple: Specialized model-sample matching for domain generalization,” in The Eleventh International Conference on Learning Representations , 2022

  20. [28]

    Mixstyle neural networks for domain generalization and adaptation,

    K. Zhou, Y . Yang, Y . Qiao, and T. Xiang, “Mixstyle neural networks for domain generalization and adaptation,” International Journal of Computer Vision, vol. 132, no. 3, pp. 822–836, 2024

  21. [29]

    Knowledge distillation-based domain-invariant representation learning for domain generalization,

    Z. Niu, J. Yuan, X. Ma, Y . Xu, J. Liu, Y .-W. Chen, R. Tong, and L. Lin, “Knowledge distillation-based domain-invariant representation learning for domain generalization,” IEEE Transactions on Multimedia , 2023

  22. [30]

    Boosting domain generalization by domain-aware knowledge distillation,

    Z. Zhang, G. Liu, F. Cai, D. Liu, and X. Fang, “Boosting domain generalization by domain-aware knowledge distillation,” Knowledge- Based Systems, vol. 280, p. 111021, 2023

  23. [31]

    Learning to generalize: Meta-learning for domain generalization,

    D. Li, Y . Yang, Y .-Z. Song, and T. Hospedales, “Learning to generalize: Meta-learning for domain generalization,” in Proceedings of the AAAI conference on artificial intelligence , vol. 32, 2018

  24. [32]

    Discriminative adversarial do- main generalization with meta-learning based cross-domain validation,

    K. Chen, D. Zhuang, and J. M. Chang, “Discriminative adversarial do- main generalization with meta-learning based cross-domain validation,” Neurocomputing, vol. 467, pp. 418–426, 2022

  25. [33]

    Learning common and specific visual prompts for domain generalization,

    A. Li, L. Zhuang, S. Fan, and S. Wang, “Learning common and specific visual prompts for domain generalization,” in Proceedings of the Asian conference on computer vision , 2022, pp. 4260–4275

  26. [34]

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

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

  27. [35]

    Consistent prompt learning for vision-language models,

    Y . Zhang and X. Tian, “Consistent prompt learning for vision-language models,” Knowledge-Based Systems, vol. 310, p. 112974, 2025

  28. [36]

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

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

  29. [37]

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

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

  30. [38]

    Clipceil: Domain generalization through clip via channel refinement and image-text alignment,

    X. Yu, S. Yoo, and Y . Lin, “Clipceil: Domain generalization through clip via channel refinement and image-text alignment,” Advances in Neural Information Processing Systems , vol. 37, pp. 4267–4294, 2024. 12

  31. [39]

    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 Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2024, pp. 5542–5552

  32. [40]

    Disentangled prompt representation for domain generalization,

    D. Cheng, Z. Xu, X. Jiang, N. Wang, D. Li, and X. Gao, “Disentangled prompt representation for domain generalization,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 595–23 604

  33. [41]

    Ensembling disentangled domain-specific prompts for domain generalization,

    F. Xu, S. Deng, T. Jia, X. Yu, and D. Chen, “Ensembling disentangled domain-specific prompts for domain generalization,” Knowledge-Based Systems, vol. 301, p. 112358, 2024

  34. [42]

    Conditional prompt learning for vision-language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Conditional prompt learning for vision-language models,” in Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , 2022, pp. 16 816– 16 825

  35. [43]

    Nlnl: Negative learning for noisy labels,

    Y . Kim, J. Yim, J. Yun, and J. Kim, “Nlnl: Negative learning for noisy labels,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 101–110

  36. [44]

    A simple framework for contrastive learning of visual representations,

    T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple framework for contrastive learning of visual representations,” in International conference on machine learning . PMLR, 2020, pp. 1597–1607

  37. [45]

    Momentum contrast for unsupervised visual representation learning,

    K. He, H. Fan, Y . Wu, S. Xie, and R. Girshick, “Momentum contrast for unsupervised visual representation learning,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 9729–9738

  38. [46]

    Learning open set network with discriminative reciprocal points,

    G. Chen, L. Qiao, Y . Shi, P. Peng, J. Li, T. Huang, S. Pu, and Y . Tian, “Learning open set network with discriminative reciprocal points,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part III 16 . Springer, 2020, pp. 507–522

  39. [47]

    Argue: Attribute-guided prompt tuning for vision-language models,

    X. Tian, S. Zou, Z. Yang, and J. Zhang, “Argue: Attribute-guided prompt tuning for vision-language models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 28 578–28 587

  40. [48]

    Clipn for zero-shot ood detection: Teaching clip to say no,

    H. Wang, Y . Li, H. Yao, and X. Li, “Clipn for zero-shot ood detection: Teaching clip to say no,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 1802–1812

  41. [49]

    Learning transferable negative prompts for out-of-distribution detection,

    T. Li, G. Pang, X. Bai, W. Miao, and J. Zheng, “Learning transferable negative prompts for out-of-distribution detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 17 584–17 594

  42. [50]

    Semi- supervised learning with pseudo-negative labels for image classifica- tion,

    H. Xu, H. Xiao, H. Hao, L. Dong, X. Qiu, and C. Peng, “Semi- supervised learning with pseudo-negative labels for image classifica- tion,” Knowledge-Based Systems, vol. 260, p. 110166, 2023

  43. [51]

    Vision-language models are strong noisy label detectors,

    T. Wei, H.-T. Li, C.-S. Li, J.-X. Shi, Y .-F. Li, and M.-L. Zhang, “Vision-language models are strong noisy label detectors,” in Advances in Neural Information Processing Systems , A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang, Eds., vol. 37...

  44. [52]

    Deeper, broader and artier domain generalization,

    D. Li, Y . Yang, Y .-Z. Song, and T. M. Hospedales, “Deeper, broader and artier domain generalization,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 5542–5550

  45. [53]

    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 Proceedings of the IEEE International Conference on Computer Vision , 2013, pp. 1657–1664

  46. [54]

    Deep hashing network for unsupervised domain adaptation,

    H. Venkateswara, J. Eusebio, S. Chakraborty, and S. Panchanathan, “Deep hashing network for unsupervised domain adaptation,” in Pro- ceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 5018–5027

  47. [55]

    Recognition in terra incognita,

    S. Beery, G. Van Horn, and P. Perona, “Recognition in terra incognita,” in Proceedings of the European conference on computer vision (ECCV) , 2018, pp. 456–473

  48. [56]

    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 Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 1406– 1415

  49. [57]

    In search of lost domain generalization,

    I. Gulrajani and D. Lopez-Paz, “In search of lost domain generalization,” in International Conference on Learning Representations , 2020

  50. [58]

    Swad: Domain generalization by seeking flat minima,

    J. Cha, S. Chun, K. Lee, H.-C. Cho, S. Park, Y . Lee, and S. Park, “Swad: Domain generalization by seeking flat minima,” Advances in Neural Information Processing Systems , vol. 34, pp. 22 405–22 418, 2021

  51. [59]

    Exploring visual prompts for adapting large-scale models,

    H. Bahng, A. Jahanian, S. Sankaranarayanan, and P. Isola, “Exploring visual prompts for adapting large-scale models,” 2022. [Online]. Available: https://arxiv.org/abs/2203.17274

  52. [60]

    Visual prompt tuning,

    M. Jia, L. Tang, B.-C. Chen, C. Cardie, S. Belongie, B. Hariharan, and S.-N. Lim, “Visual prompt tuning,” in European Conference on Computer Vision. Springer, 2022, pp. 709–727

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.