Pith. sign in

REVIEW 5 major objections 6 minor 75 references

A Simple and Efficient Baseline for Zero-Shot Generative Classification

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read GDC makes zero-shot diffusion classification practical by fitting per-class Gaussians to DINOv2 embeddings of generated images, reaching 71.4% on ImageNet at 0.03 seconds per image.

desk verdict GDC makes diffusion-based zero-shot classification practical, but the accuracy gain over prior diffusion classifiers is confounded by the simultaneous switch to DINOv2 and SDXL-turbo, and the paper never runs the obvious same-embedding prototype control. read the letter →

arxiv 2412.12594 v1 pith:3VW3AMO3 submitted 2024-12-17 cs.CV

classification cs.CV
keywords zero-shotclassificationdiffusionmodelsGaussianmixturemodelDINOv2generativeclassifiersImageNetefficiencyBayesian
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 a zero-shot classifier, the Gaussian Diffusion Classifier (GDC), that avoids the usual per-image cost of diffusion-based classification. During a one-time preparation phase, it prompts a text-to-image diffusion model to generate reference images for each class, embeds them with DINOv2, and fits one Gaussian per class in embedding space. At test time, classifying an image is a single encoder forward pass plus a likelihood evaluation, so a 1,000-class image that took earlier diffusion classifiers about 1,100 seconds takes about 0.03 seconds. The paper reports that this simple replacement also improves ImageNet accuracy from 61.40% to 71.44% over previous diffusion-based zero-shot classifiers, and that accuracy rises as the underlying diffusion model gets stronger. If correct, GDC is the first zero-shot diffusion-based classifier that is both competitively accurate and fast enough for practical use.

What carries the argument

The central object is a Gaussian Mixture Model built in the embedding space of DINOv2, where each class component is fit to DINOv2 embeddings of images generated by the diffusion model from class-name prompts. The load-bearing identity is the Gaussian log-likelihood with a Cholesky-decomposed regularized precision matrix, $\hat{\Sigma}_i^{-1} = (\Sigma_i + \epsilon I)^{-1} = LL^*$, used to compute $p(e|y_i)$; Bayes' rule then yields $p(y_i|e)$ and the argmax label. The two-phase design moves all reverse-diffusion cost into the preparation phase, so per-image test cost is one encoder forward pass and $k$ Gaussian density evaluations.

What would settle it

On ImageNet, fit each class's Gaussian to 240 generated references as in GDC, then refit it to 240 real training images per class while keeping the same DINOv2 encoder and test set; if the real-fit classifier's accuracy does not clearly exceed the generated-fit one, the distribution-match premise survives, and if it does, the central assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a diffusion model can be turned into a zero-shot classifier without ever running the diffusion denoising loop at test time. Instead of estimating per-class denoising losses for each test image, GDC generates reference images for each class caption in advance, embeds them with DINOv2, and models each class's embedding distribution as a Gaussian. Classification then follows Bayes' rule: assign a test image to the class whose Gaussian gives the highest likelihood, with class priors folded in. The paper argues that this preserves the semantic knowledge of the diffusion model while replacing thousands of reverse passes per image with one forward pass of an image encoder, and it reports that the resulting classifier beats previous loss-based diffusion classifiers by more than ten points on ImageNet while running more than thirty thousand times faster.

Load-bearing premise

The method assumes DINOv2 embeddings of diffusion-generated reference images follow the same class-conditional Gaussian distribution as real test images, and that a 1536-dimensional covariance per class can be estimated from 240 generated samples; the paper's one-shot experiment shows replacing one generated image with a real one adds 4.8 points on ImageNet, indicating the match is imperfect.

Editorial extensions

If this is right

  • Zero-shot classification with diffusion models shifts from a per-image cost of thousands of reverse diffusion passes to a prepaid cost: for a fixed label set, the expensive generation happens once, so adding test images is nearly free.
  • Accuracy tracks generation quality: as text-to-image models improve, GDC's zero-shot accuracy should keep rising without any change to the classifier, making ImageNet accuracy a potential proxy for generation quality.
  • The classifier outputs posterior probabilities that sum to one, enabling uncertainty quantification and risk assessment for downstream applications.
  • With one real image per class, GDC exceeds CLIP's one-shot accuracy on ImageNet (76.2 vs 75.2), suggesting the framework can absorb small amounts of labeled data when available.
  • Because no training data is used, GDC can be applied to a new label set by re-running only the preparation phase for those class names.

Reading between the lines

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

  • A natural extension the paper leaves untested is open-set or anomaly detection: because every class is a Gaussian, out-of-distribution test embeddings should have low likelihood under all classes, though the paper does not evaluate this.
  • Since only the encoder and Gaussian fits matter at test time, swapping in a stronger image encoder should improve accuracy without retraining the diffusion model, a prediction that follows from the method's architecture rather than from the paper's experiments.
  • The scaling result suggests ImageNet zero-shot accuracy could become an inexpensive, human-preference-free proxy for text-to-image generation quality, a direction the paper mentions only as future work.
  • The error analysis implies that prompt diversity, not just diffusion model quality, is the main lever for closing the remaining gap: errors concentrate on hidden, coexisting, or unusually viewed objects that the default prompt templates underrepresent.
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 / 6 minor

Summary. The paper proposes the Gaussian Diffusion Classifier (GDC), a zero-shot generative classifier that first uses a pretrained text-to-image diffusion model (SDXL-turbo) to generate N reference images for each class, embeds them with DINOv2, and fits a per-class full-covariance Gaussian distribution. At test time, a query image is embedded with DINOv2 and classified by maximum Gaussian likelihood. The paper reports large accuracy gains over prior loss-based diffusion classifiers (e.g., 71.4% vs. 61.4% on ImageNet), massive speedups (0.03 s vs. 1133 s per image), competitive accuracy across several benchmarks, and a one-shot variant that outperforms CLIP. The central claims are that GDC is the first zero-shot diffusion-based classifier to combine competitive accuracy with practical efficiency and that it self-improves with stronger diffusion backbones.

Significance. The efficiency contribution is substantial and credible: replacing per-image diffusion reverse inference with a small set of pre-generated reference clusters reduces deployment cost by several orders of magnitude while providing a probabilistic output. The proposed procedure is simple and easy to reproduce, and the monotonic improvement with stronger diffusion models is an interesting scaling property. However, the accuracy claims rest on an uncontrolled comparison: GDC differs from the baselines in both the diffusion backbone and the image encoder, and the paper does not ablate the Gaussian decision rule against a nearest-class-mean or diagonal-covariance classifier on the same embeddings. If those ablations were added and supported the method, the paper would be a strong baseline contribution; as written, the headline accuracy gain is not yet attributed to the proposed classifier model.

major comments (5)
  1. [§5, Table 1] The headline comparison is not controlled. GDC uses SDXL-turbo for generation and DINOv2 for embedding, while the two baseline diffusion classifiers use an older Stable Diffusion backbone and do not use DINOv2. The reported 10-point ImageNet gain could therefore come from the encoder or backbone rather than from the Gaussian classifier. The authors should include a nearest-class-mean (or diagonal-covariance) classifier on the very same DINOv2 embeddings of the same generated reference images. If that baseline approaches 71.4%, then the full-covariance Gaussian machinery is not the source of the improvement and the claim that GDC is the first competitive zero-shot diffusion classifier is overstated.
  2. [§5, Appendix B] The default regularization value epsilon=1e-8 is chosen after inspecting test-set accuracies (Table 5), and N and the prompt templates are similarly selected from benchmark results (Figure 3 and Table 4). Because the paper advertises zero-shot classification with no training data, tuning these hyperparameters on the test set biases the reported numbers and weakens the zero-shot claim. The authors should fix hyperparameters a priori, report a development-split selection, or systematically demonstrate that the accuracy is insensitive across a wide range of values for each hyperparameter.
  3. [§5, Table 3] The one-shot experiment replaces one generated image per class with a real training image and improves ImageNet accuracy from 71.4% to 76.2%. This is direct evidence that the DINOv2 embeddings of SDXL-turbo reference images do not occupy the same class-conditional distribution as real test images. The paper should analyze this distribution gap explicitly (e.g., by reporting performance versus the number of real images, or by evaluating a simple domain-gap measure) and should qualify the zero-shot claims accordingly, since the one-shot CLIP comparison does not establish zero-shot competitiveness.
  4. [§3, Eq. (1), Appendix B] With N=240 samples per class and d=1536, the empirical covariance is singular, and the regularized precision matrix in Eq. (1) relies on epsilon=1e-8 to be invertible; epsilon=1e-10 yields NaN (Appendix B). The Gaussian log-likelihood in Eq. (3) is then strongly affected by the projection of test embeddings onto the null space of the sample covariance. The paper should report the effective rank of the regularized covariance, compare against diagonal or shrunk covariance estimators on the same embeddings, and demonstrate that the full-covariance model is actually beneficial relative to simpler estimators.
  5. [§5] All reported accuracies are point estimates from a single run of stochastic reference-image generation. Because SDXL-turbo is non-deterministic, the clusters and hence the resulting accuracy are random. The paper should report the mean and standard deviation over at least three seeds for the main benchmarks (at least ImageNet and CIFAR-100) to establish that the comparisons in Tables 1 and 4 are not noise.
minor comments (6)
  1. [Algorithm 1] The pseudocode uses 'k' both as class index and sample index in the preparation phase, and 'n' both as number of reference images and number of test samples in the classification phase; please rename to avoid ambiguity.
  2. [§4, Eq. (3)] The mean vectors are denoted 'u_i' in the Gaussian density but 'µ_i' in the GMM construction; use consistent notation throughout.
  3. [§5, Table 2] The column 'Classification Total Time' mixes preparation and deployment time; label the rows more clearly (e.g., 'preparation', 'test-set inference', 'total') so that the 0.03 s single-image time is not confused with the total preparation time.
  4. [Throughout] There are multiple typos, including 'probility', 'caculate', 'ther are', 'As code relase problem', and 'CGC' (should be GDC) in the discussion of Table 1; these should be corrected.
  5. [§3, Figure 1] The caption says CIFAR-10 while the text refers to CIFAR-100 and ImageNet; the figure and caption should be aligned.
  6. [Appendix C] The supplementary normality test reports p-values per principal component; please clarify how the 60.01% figure across classes and components is aggregated (e.g., fraction of tests with p>0.05 over all class-component pairs).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: GDC is a straightforward Bayesian likelihood classifier whose Gaussian parameters are fit exclusively to generated reference images and whose accuracy is measured on held-out test data; self-citations are peripheral and not load-bearing.

full rationale

The paper's derivation chain is self-contained, and its headline numbers are external-benchmark measurements rather than outputs forced by its own construction. GDC classifies a test image by embedding it with DINOv2 and computing Gaussian class-conditional densities p(e|yi), with each class Gaussian (mean vector and covariance matrix) estimated exclusively from text-to-image-generated reference images (Algorithm 1 and Section 4, Preparation Phase). No parameter is fit to test labels, so the reported accuracies (e.g., 71.44% on ImageNet) are genuine held-out evaluations: the 'prediction' does not reduce to the fit by construction, and the 10-point gain over Li's DC is a comparison against published external baselines. The Gaussian ansatz is motivated by Shapiro-Wilk normality tests on DINOv2 features in Section 3 and Appendix C, which is an external statistical test, not a self-citation. The paper's several self-citations ([1], [31], [48], [49], [50], [59], [74]) appear only in related-work and diffusion-improvement discussions (e.g., 'obtaining better initial noise [17, 50, 74], optimizing the sampling strategy [1, 32, 33, 59]'), and none is load-bearing for the central accuracy, efficiency, or first-competitive-classifier claims; the monotonic-improvement claim rests on the paper's own Figure 4 experiment, not on a cited result. The skeptic's concerns are real but are not circularity: the missing same-backbone nearest-class-mean ablation is a comparison-validity gap; the epsilon=1e-8 / N=240 / prompt-template choices are test-set hyperparameter tuning (a mild benchmark-fitting concern that could bias reported numbers optimistically, but does not make the prediction statistically forced by construction); and the one-shot gain in Table 3 (replacing one generated image with a real image improves ImageNet accuracy from 71.4 to 76.2) demonstrates a generated-versus-real distribution gap, which is an empirical weakness and the opposite of circularity, since the classifier is being tested on data it never saw. No step in the paper's own equations or citations reduces a claimed result to its inputs.

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

The method rests on pretrained models and design choices: N, epsilon, and prompt templates are tuned rather than derived. The Gaussian distribution match between generated references and real test images is the load-bearing assumption. GDC introduces no new physical entities or latent quantities; it is a composition of existing models and statistical estimation.

free parameters (3)
  • Number of reference images per class N = 240
    Chosen after ablation; larger N improves accuracy but costs preparation time (Figure 3). This is a hyperparameter tuned to datasets, not derived from theory.
  • Covariance regularization epsilon = 1e-8
    Selected from ablation on test sets (Appendix B, Table 5); lower values give higher accuracy until numerical instability (NaN at 1e-10). This is tuned to benchmark performance.
  • Prompt augmentation template set = eight templates
    Manual design choice; ablation without augmentation drops accuracy by 2.9 to 9.6 points (Table 4). It is a hand-picked component, not an automatic fit.
assumptions (3)
  • domain assumption DINOv2 embeddings of each class's generated reference images follow a multivariate Gaussian distribution
    Section 3 and Appendix C test only marginal Gaussianity; only about 60% of principal components pass Shapiro-Wilk at p>0.05, yet the method relies on a full covariance Gaussian likelihood.
  • domain assumption Text-to-image generated reference images are representative of real test images of the same class in DINOv2 feature space
    Section 4 builds the GMM entirely from SDXL-turbo generations; if generated images do not cover the real distribution, the Gaussian likelihood will systematically misclassify. The one-shot experiment suggests real images help, implying the zero-shot fit is imperfect.
  • domain assumption DINOv2 provides a discriminative embedding space where semantic classes cluster
    The method's accuracy depends on DINOv2 separating categories; the paper treats DINOv2 as a fixed feature extractor without evaluating alternative encoders or a linear-probe baseline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Simple and Efficient Baseline for Zero-Shot Generative Classification." pith.science (2026). https://pith.science/paper/3VW3AMO3

@misc{pith2026241212594,
  author       = {Pith},
  title        = {Pith review of: A Simple and Efficient Baseline for Zero-Shot Generative Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3VW3AMO3}},
  note         = {Machine review of arXiv:2412.12594}
}
read the original abstract

Large diffusion models have become mainstream generative models in both academic studies and industrial AIGC applications. Recently, a number of works further explored how to employ the power of large diffusion models as zero-shot classifiers. While recent zero-shot diffusion-based classifiers have made performance advancement on benchmark datasets, they still suffered badly from extremely slow classification speed (e.g., ~1000 seconds per classifying single image on ImageNet). The extremely slow classification speed strongly prohibits existing zero-shot diffusion-based classifiers from practical applications. In this paper, we propose an embarrassingly simple and efficient zero-shot Gaussian Diffusion Classifiers (GDC) via pretrained text-to-image diffusion models and DINOv2. The proposed GDC can not only significantly surpass previous zero-shot diffusion-based classifiers by over 10 points (61.40% - 71.44%) on ImageNet, but also accelerate more than 30000 times (1000 - 0.03 seconds) classifying a single image on ImageNet. Additionally, it provides probability interpretation of the results. Our extensive experiments further demonstrate that GDC can achieve highly competitive zero-shot classification performance over various datasets and can promisingly self-improve with stronger diffusion models. To the best of our knowledge, the proposed GDC is the first zero-shot diffusionbased classifier that exhibits both competitive accuracy and practical efficiency.

Figures

Figures reproduced from arXiv: 2412.12594 by the authors.

Figure 1
Figure 1. We visualize the distribution of image features for random [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The overview of our Gaussian Diffusion Classifiers (GDC), which consists of two phases: 1) Preparation Phase and 2) Gaussian [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The test accuracy of GDC with various choices of [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The classification accuracy of GDC monotonically increases [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: Some visual results misclassified by Li’s DC are correctly Crane … Trench coat … [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Correct Cases and Erroneous Cases. We show examples [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: We visualize the distribution of image features for random [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

75 extracted references · 51 canonical work pages

  1. [1]

    Zigzag diffusion sampling: The path to success is zigzag, 2024

    Lichen Bai, Shitong Shao, Zikai Zhou, Zipeng Qi, Zhiqiang Xu, Haoyi Xiong, and Zeke Xie. Zigzag diffusion sampling: The path to success is zigzag, 2024. 6

  2. [2]

    Commandant Benoit. Note sur une méthode de résolution des équations normales provenant de l’application de la méth- ode des moindres carrés à un système d’équations linéaires en nombre inférieur à celui des inconnues (procédé du com- mandant cholesky). Bulletin géodésique, 2(1):67–77, 1924. 13

  3. [3]

    Stable video diffusion: Scaling latent video diffusion models to large datasets

    Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram V oleti, Adam Letts, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127, 2023. 1

  4. [4]

    Food-101–mining discriminative components with random forests

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101–mining discriminative components with random forests. In Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Pro- ceedings, Part VI 13, pages 446–461. Springer, 2014. 5, 12

  5. [5]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Sub- biah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020. 2

  6. [6]

    Gen- erating visual representations for zero-shot classification

    Maxime Bucher, Stéphane Herbin, and Frédéric Jurie. Gen- erating visual representations for zero-shot classification. In Proceedings of the IEEE International Conference on Com- puter Vision Workshops, pages 2666–2673, 2017. 2

  7. [7]

    Peekaboo: Text to image diffusion models are zero-shot segmentors

    Ryan Burgert, Kanchana Ranasinghe, Xiang Li, and Michael S Ryoo. Peekaboo: Text to image diffusion models are zero-shot segmentors. arXiv preprint arXiv:2211.13224,

  8. [8]

    Robust clas- sification via a single diffusion model

    Huanran Chen, Yinpeng Dong, Zhengyi Wang, Xiao Yang, Chengqi Duan, Hang Su, and Jun Zhu. Robust clas- sification via a single diffusion model. arXiv preprint arXiv:2305.15241, 2023. 2

Show all 75 references
  1. [9]

    Cimpoi, S

    M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, , and A. Vedaldi. Describing textures in the wild. In Proceedings of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2014. 5, 12

  2. [10]

    Text-to-image diffusion mod- els are zero-shot classifiers

    Kevin Clark and Priyank Jaini. Text-to-image diffusion mod- els are zero-shot classifiers. In ICLR 2023 Workshop on Mathematical and Empirical Understanding of Foundation Models, 2023. 1, 3, 5, 6

  3. [11]

    An analysis of single-layer networks in unsupervised feature learning

    Adam Coates, Andrew Ng, and Honglak Lee. An analysis of single-layer networks in unsupervised feature learning. In Proceedings of the fourteenth international conference on artificial intelligence and statistics, pages 215–223. JMLR Workshop and Conference Proceedings, 2011. 5, 12

  4. [12]

    Gan- bert: Generative adversarial learning for robust text classifica- tion with a bunch of labeled examples

    Danilo Croce, Giuseppe Castellucci, and Roberto Basili. Gan- bert: Generative adversarial learning for robust text classifica- tion with a bunch of labeled examples. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2114–2119, 2020. 2

  5. [13]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009. 1, 5, 12

  6. [14]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021. 1, 2

  7. [15]

    Class prior estimation from positive and unlabeled data

    Marthinus Christoffel Du Plessis and Masashi Sugiyama. Class prior estimation from positive and unlabeled data. IE- ICE TRANSACTIONS on Information and Systems , 97(5): 1358–1362, 2014. 4

  8. [16]

    One-shot learn- ing of object categories

    Li Fei-Fei, Robert Fergus, and Pietro Perona. One-shot learn- ing of object categories. IEEE transactions on pattern analy- sis and machine intelligence, 28(4):594–611, 2006. 5, 12

  9. [17]

    Initno: Boosting text-to-image diffu- sion models via initial noise optimization

    Xiefan Guo, Jinlin Liu, Miaomiao Cui, Jiankai Li, Hongyu Yang, and Di Huang. Initno: Boosting text-to-image diffu- sion models via initial noise optimization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9380–9389, 2024. 6

  10. [18]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 16000– 16009, 2022. 2

  11. [19]

    Learning deep representations by mutual information estimation and maximization

    R Devon Hjelm, Alex Fedorov, Samuel Lavoie-Marchildon, Karan Grewal, Phil Bachman, Adam Trischler, and Yoshua Bengio. Learning deep representations by mutual information estimation and maximization. In International Conference on Learning Representations, 2018. 2

  12. [20]

    Denoising diffu- sion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 1

  13. [21]

    Imagen video: High definition video generation with diffusion models.arXiv preprint arXiv:2210.02303, 2022

    Jonathan Ho, William Chan, Chitwan Saharia, Jay Whang, Ruiqi Gao, Alexey Gritsenko, Diederik P Kingma, Ben Poole, Mohammad Norouzi, David J Fleet, et al. Imagen video: High definition video generation with diffusion models.arXiv preprint arXiv:2210.02303, 2022. 1

  14. [22]

    Intriguing properties of generative classifiers

    Priyank Jaini, Kevin Clark, and Robert Geirhos. Intriguing properties of generative classifiers. In The Twelfth Interna- tional Conference on Learning Representations , 2023. 1, 2

  15. [23]

    Learning discriminative latent attributes for zero- shot classification

    Huajie Jiang, Ruiping Wang, Shiguang Shan, Yi Yang, and Xilin Chen. Learning discriminative latent attributes for zero- shot classification. In Proceedings of the IEEE International Conference on Computer Vision, pages 4223–4232, 2017. 2

  16. [24]

    Diffwave: A versatile diffusion model for audio synthesis

    Zhifeng Kong, Wei Ping, Jiaji Huang, Kexin Zhao, and Bryan Catanzaro. Diffwave: A versatile diffusion model for audio synthesis. In International Conference on Learning Repre- sentations, 2020. 1

  17. [25]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky and Geoffrey Hinton. Learning multiple layers of features from tiny images. 2009. 5, 12

  18. [26]

    Neural network classification and prior class probabilities

    Steve Lawrence, Ian Burns, Andrew Back, Ah Chung Tsoi, and C Lee Giles. Neural network classification and prior class probabilities. In Neural networks: tricks of the trade, pages 299–313. Springer, 2002. 4

  19. [27]

    Robust inference via generative classifiers for handling noisy labels

    Kimin Lee, Sukmin Yun, Kibok Lee, Honglak Lee, Bo Li, and Jinwoo Shin. Robust inference via generative classifiers for handling noisy labels. In International conference on machine learning, pages 3763–3772. PMLR, 2019. 2

  20. [28]

    Li, Mihir Prabhudesai, Shivam Duggal, Ellis Brown, and Deepak Pathak

    Alexander C. Li, Mihir Prabhudesai, Shivam Duggal, Ellis Brown, and Deepak Pathak. Your diffusion model is secretly a zero-shot classifier. In Thirty-seventh International Confer- ence on Computer Vision, 2023. 1, 3, 5, 6

  21. [29]

    Are genera- tive classifiers more robust to adversarial attacks? In Interna- tional Conference on Machine Learning, pages 3804–3814

    Yingzhen Li, John Bradshaw, and Yash Sharma. Are genera- tive classifiers more robust to adversarial attacks? In Interna- tional Conference on Machine Learning, pages 3804–3814. PMLR, 2019. 2

  22. [30]

    Magic3d: High-resolution text- to-3d content creation

    Chen-Hsuan Lin, Jun Gao, Luming Tang, Towaki Takikawa, Xiaohui Zeng, Xun Huang, Karsten Kreis, Sanja Fidler, Ming- Yu Liu, and Tsung-Yi Lin. Magic3d: High-resolution text- to-3d content creation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognit...

  23. [31]

    Alignment of diffusion models: Fundamentals, challenges, and future

    Buhua Liu, Shitong Shao, Bao Li, Lichen Bai, Zhiqiang Xu, Haoyi Xiong, James Kwok, Sumi Helal, and Zeke Xie. Alignment of diffusion models: Fundamentals, challenges, and future. arXiv preprint arXiv:2409.07253, 2024. 6

  24. [32]

    Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, and Jun Zhu. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. Advances in Neural Information Processing Systems, 35:5775–5787, 2022. 6

  25. [33]

    Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, and Jun Zhu. Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models. arXiv preprint arXiv:2211.01095, 2022. 6

  26. [34]

    Generative classifiers as a basis for trustwor- thy image classification

    Radek Mackowiak, Lynton Ardizzone, Ullrich Kothe, and Carsten Rother. Generative classifiers as a basis for trustwor- thy image classification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 2971–2981, 2021. 2

  27. [35]

    Costa: Co-occurrence statistics for zero-shot classification

    Thomas Mensink, Efstratios Gavves, and Cees GM Snoek. Costa: Co-occurrence statistics for zero-shot classification. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2441–2448, 2014. 2

  28. [36]

    On discriminative vs

    Andrew Ng and Michael Jordan. On discriminative vs. genera- tive classifiers: A comparison of logistic regression and naive bayes. Advances in neural information processing systems, 14, 2001. 2

  29. [37]

    Glide: Towards photorealistic image generation and editing with text-guided diffusion models

    Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew, Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with text-guided diffusion models. arXiv preprint arXiv:2112.10741, 2021. 1

  30. [38]

    Glide: Towards photorealistic image gen- eration and editing with text-guided diffusion models

    Alexander Quinn Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob Mcgrew, Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image gen- eration and editing with text-guided diffusion models. In International Conference on Machine Learning, p...

  31. [39]

    Automated flower classification over a large number of classes

    Maria-Elena Nilsback and Andrew Zisserman. Automated flower classification over a large number of classes. In 2008 Sixth Indian conference on computer vision, graphics & image processing, pages 722–729. IEEE, 2008. 5, 12

  32. [40]

    Per- vasive label errors in test sets destabilize machine learning benchmarks

    Curtis G Northcutt, Anish Athalye, and Jonas Mueller. Per- vasive label errors in test sets destabilize machine learning benchmarks. In Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1), 2021. 7

  33. [41]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 2, 3, 5

  34. [42]

    Watt: Weight average test-time adaption of clip

    David Osowiechi, Mehrdad Noori, Gustavo Adolfo Vargas Hakim, Moslem Yazdanpanah, Ali Bahri, Milad Cheragha- likhani, Sahar Dastani, Farzad Beizaee, Ismail Ben Ayed, and Christian Desrosiers. Watt: Weight average test-time adaption of clip. arXiv preprint arXiv:2406.13875, 2024. 3

  35. [43]

    Cats and dogs

    Omkar M Parkhi, Andrea Vedaldi, Andrew Zisserman, and CV Jawahar. Cats and dogs. In 2012 IEEE conference on computer vision and pattern recognition, pages 3498–3505. IEEE, 2012. 5, 12

  36. [44]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 4195–4205,

  37. [45]

    Sdxl: Improving latent diffusion models for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023. 1

  38. [46]

    Dreamfusion: Text-to-3d using 2d diffusion

    Ben Poole, Ajay Jain, Jonathan T Barron, and Ben Mildenhall. Dreamfusion: Text-to-3d using 2d diffusion. In The Eleventh International Conference on Learning Representations, 2022. 1

  39. [47]

    Aligning text-to-image diffusion models with reward backpropagation

    Mihir Prabhudesai, Anirudh Goyal, Deepak Pathak, and Katerina Fragkiadaki. Aligning text-to-image diffusion models with reward backpropagation. arXiv preprint arXiv:2310.03739, 2023. 6

  40. [48]

    Layered rendering diffusion model for zero-shot guided image synthesis

    Zipeng Qi, Guoxi Huang, Zebin Huang, Qin Guo, Jinwen Chen, Junyu Han, Jian Wang, Gang Zhang, Lufei Liu, Errui Ding, et al. Layered rendering diffusion model for zero-shot guided image synthesis. arXiv preprint arXiv:2311.18435,

  41. [49]

    Difftalker: Co-driven audio-image diffusion for talking faces via intermediate landmarks

    Zipeng Qi, Xulong Zhang, Ning Cheng, Jing Xiao, and Jian- zong Wang. Difftalker: Co-driven audio-image diffusion for talking faces via intermediate landmarks. arXiv preprint arXiv:2309.07509, 2023. 1

  42. [50]

    Not all noises are created equally: Diffusion noise selection and optimization

    Zipeng Qi, Lichen Bai, Haoyi Xiong, et al. Not all noises are created equally: Diffusion noise selection and optimization. arXiv preprint arXiv:2407.14041, 2024. 1, 6

  43. [51]

    Language models are unsuper- vised multitask learners

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language models are unsuper- vised multitask learners. OpenAI blog, 1(8):9, 2019. 2

  44. [52]

    Learning transferable visual models from natural language supervi- sion

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

  45. [53]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In International Confer- ence on Machine Learning, pages 8821–8831. PMLR, 2021. 2

  46. [54]

    Hierarchical text-conditional image genera- tion with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image genera- tion with clip latents. arXiv preprint arXiv:2204.06125, 1(2): 3, 2022. 2

  47. [55]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 1, 2, 6

  48. [56]

    Pho- torealistic text-to-image diffusion models with deep language understanding

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Pho- torealistic text-to-image diffusion models with deep language understanding. Advances in neural informati...

  49. [57]

    Adversarial diffusion distillation

    Axel Sauer, Dominik Lorenz, Andreas Blattmann, and Robin Rombach. Adversarial diffusion distillation. arXiv preprint arXiv:2311.17042, 2023. 1

  50. [58]

    Uncertainty-aware deep classifiers using generative models

    Murat Sensoy, Lance Kaplan, Federico Cerutti, and Maryam Saleki. Uncertainty-aware deep classifiers using generative models. In Proceedings of the AAAI conference on artificial intelligence, pages 5620–5627, 2020. 2

  51. [59]

    Iv-mixed sampler: Leveraging image diffu- sion models for enhanced video synthesis

    Shitong Shao, Zikai Zhou, Lichen Bai, Haoyi Xiong, and Zeke Xie. Iv-mixed sampler: Leveraging image diffu- sion models for enhanced video synthesis. arXiv preprint arXiv:2410.04171, 2024. 6

  52. [60]

    An analysis of variance test for normality (complete samples)

    Samuel Sanford Shapiro and Martin B Wilk. An analysis of variance test for normality (complete samples). Biometrika, 52(3-4):591–611, 1965. 3, 12, 13

  53. [61]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In International confer- ence on machine learning, pages 2256–2265. PMLR, 2015. 1

  54. [62]

    Generative modeling by estimating gradients of the data distribution

    Yang Song and Stefano Ermon. Generative modeling by estimating gradients of the data distribution. Advances in neural information processing systems, 32, 2019. 1

  55. [63]

    Class- incremental learning with generative classifiers

    Gido M Van De Ven, Zhe Li, and Andreas S Tolias. Class- incremental learning with generative classifiers. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3611–3620, 2021. 2

  56. [64]

    Diffusion model align- ment using direct preference optimization

    Bram Wallace, Meihua Dang, Rafael Rafailov, Linqi Zhou, Aaron Lou, Senthil Purushwalkam, Stefano Ermon, Caiming Xiong, Shafiq Joty, and Nikhil Naik. Diffusion model align- ment using direct preference optimization. In Proceedings of the IEEE/CVF Conference on Computer Vision a...

  57. [65]

    A survey of zero-shot learning: Settings, methods, and ap- plications

    Wei Wang, Vincent W Zheng, Han Yu, and Chunyan Miao. A survey of zero-shot learning: Settings, methods, and ap- plications. ACM Transactions on Intelligent Systems and Technology (TIST), 10(2):1–37, 2019. 2

  58. [66]

    Finetuned language models are zero-shot learners

    Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. Finetuned language models are zero-shot learners. arXiv preprint arXiv:2109.01652, 2021. 2

  59. [67]

    Robust fine-tuning of zero-shot models

    Mitchell Wortsman, Gabriel Ilharco, Jong Wook Kim, Mike Li, Simon Kornblith, Rebecca Roelofs, Raphael Gon- tijo Lopes, Hannaneh Hajishirzi, Ali Farhadi, Hongseok Namkoong, et al. Robust fine-tuning of zero-shot models. In Proceedings of the IEEE/CVF Conference on Computer Visi...

  60. [68]

    Latent embed- dings for zero-shot classification

    Yongqin Xian, Zeynep Akata, Gaurav Sharma, Quynh Nguyen, Matthias Hein, and Bernt Schiele. Latent embed- dings for zero-shot classification. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 69–77, 2016. 2

  61. [69]

    Dream3d: Zero- shot text-to-3d synthesis using 3d shape prior and text-to- image diffusion models

    Jiale Xu, Xintao Wang, Weihao Cheng, Yan-Pei Cao, Ying Shan, Xiaohu Qie, and Shenghua Gao. Dream3d: Zero- shot text-to-3d synthesis using 3d shape prior and text-to- image diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , ...

  62. [70]

    Diffsound: Discrete diffusion model for text-to-sound generation

    Dongchao Yang, Jianwei Yu, Helin Wang, Wen Wang, Chao Weng, Yuexian Zou, and Dong Yu. Diffsound: Discrete diffusion model for text-to-sound generation. IEEE/ACM Transactions on Audio, Speech, and Language Processing ,

  63. [71]

    Diffusion models: A comprehensive survey of methods and applications

    Ling Yang, Zhilong Zhang, Yang Song, Shenda Hong, Run- sheng Xu, Yue Zhao, Wentao Zhang, Bin Cui, and Ming- Hsuan Yang. Diffusion models: A comprehensive survey of methods and applications. ACM Computing Surveys, 56(4): 1–39, 2023. 1, 2

  64. [72]

    Zero-shot classification with discriminative semantic representation learning

    Meng Ye and Yuhong Guo. Zero-shot classification with discriminative semantic representation learning. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 7140–7148, 2017. 2

  65. [73]

    Revisiting discriminative vs

    Chenyu Zheng, Guoqiang Wu, Fan Bao, Yue Cao, Chongxuan Li, and Jun Zhu. Revisiting discriminative vs. generative clas- sifiers: Theory and implications. In International Conference on Machine Learning, 2023. 2

  66. [74]

    Golden noise for diffusion models: A learning framework

    Zikai Zhou, Shitong Shao, Lichen Bai, Zhiqiang Xu, Bo Han, and Zeke Xie. Golden noise for diffusion models: A learning framework. arXiv preprint arXiv:2411.09502, 2024. 6

  67. [75]

    Score-based generative classifiers

    Roland S Zimmermann, Lukas Schott, Yang Song, Ben- jamin A Dunn, and David A Klindt. Score-based generative classifiers. arXiv preprint arXiv:2110.00473, 2021. 2 A Simple and Efficient Baseline for Zero-Shot Generative Classification Supplementary Material A. Experimental Sett...

Pith tools

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