REVIEW 3 major objections 4 minor 114 references
The GAN is dead; long live the GAN! A Modern GAN Baseline
T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash
Pith's one-line read With the right regularized loss, GAN training needs no tricks: the minimalist R3GAN beats StyleGAN2 and matches diffusion models in one forward pass.
desk verdict A credible, well-executed modern GAN baseline with a real gap between the local convergence theory and the R1+R2 recipe that actually stabilizes training; worth serious review. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the regularized relativistic pairing loss (RpGAN), which compares each fake sample against a real sample rather than against a fixed decision boundary: $L(\theta,\psi)=\mathbb{E}_{z\sim p_z, x\sim p_D}[f(D_\psi(G_\theta(z))-D_\psi(x))] + \frac{\gamma}{2}\mathbb{E}_{x\sim p_D}[\|\nabla_x D_\psi\|^2] + \frac{\gamma}{2}\mathbb{E}_{x\sim p_\theta}[\|\nabla_x D_\psi\|^2]$, where $f(t)=-\log(1+e^{-t})$, plus the decision to apply gradient penalties on both the real and fake distributions. The convergence argument follows the established technique of linearizing the gradient vector field $v(\theta,\psi)=(-\nabla_\theta L,\nabla_\psi L)$ at the equilibrium $(\theta^*, \psi^*)$ where $p_{\theta^*}=p_D$ and $D_{\psi^*}$ is constant on the data support: if the Jacobian's eigenvalues all have negative real part, gradient descent converges locally at a linear rate. The paper shows that for unregularized RpGAN the Jacobian has purely imaginary eigenvalues (a concrete DiracRpGAN example runs in circles), while the penalties shift the spectrum left, yielding the local-convergence theorem. On the architecture side, the mechanism that carries the empirical result is the pairing of this stable loss with a modern symmetric ResNet-style backbone built from bilinear resampling, grouped convolutions, inverted bottlenecks, and fix-up initialization: the absence of normalization layers is compensated by careful initialization, and the absence of tricks is compensated by the well-behaved loss.
What would settle it
A direct test would be to train the exact trick-free R3GAN recipe on a held-out dataset such as LSUN bedrooms or ImageNet-256: if training diverges or collapses unless some removed trick is restored, the claim that the RpGAN+R1+R2 loss alone provides sufficient stability would fail; a cheaper check is to rerun the Stacked-MNIST experiment with RpGAN+R1 alone across many seeds and a tuned learning-rate schedule, since the paper's central practical justification for needing R2 is that R1-only training diverges from typical starts.
Extended reading notes
Core claim
The central discovery is that the combination of a relativistic pairing loss and both zero-centered gradient penalties — R1 on real data and R2 on fake data — turns GAN training into a locally convergent dynamical system, and that this one change is what makes a trick-free, modern architecture viable. The paper proves that unregularized RpGAN does not generally converge: in a Dirac-style toy setting its gradient dynamics run on circles around the equilibrium, and the Jacobian there has purely imaginary eigenvalues. Adding either R1 or R2 moves those eigenvalues into the left half-plane, giving linear local convergence under the standard realizable-case assumptions (the generator can exactly match the data, and the discriminator can detect any deviation from equilibrium). With this loss alone, the authors strip StyleGAN2 down to a bare backbone — no mapping network, style injection, minibatch stddev, equalized learning rate, or path-length regularization — then modernize it with ResNet/ConvNeXt-style blocks, grouped convolutions, and an inverted bottleneck. The resulting R3GAN reports FID 2.75 on FFHQ-256 (StyleGAN2: 3.78), 1.95 on FFHQ-64 (EDM: 2.39), 1.96 on CIFAR-10, 1.27 on ImageNet-32, and 2.09 on ImageNet-64 (EDM: 2.23), all with a single function evaluation, and it recovers all 1000 modes of Stacked-MNIST with the lowest KL divergence among the GANs tested.
Load-bearing premise
The convergence proof only applies near an equilibrium where the generator already matches the data distribution almost exactly and the discriminator is expressive enough to detect any small divergence from it — a condition that cannot be verified for the finite networks actually trained, and the paper's own Stacked-MNIST runs show that the theoretically-covered R1-only variant still diverges from random initialization.
Editorial extensions
If this is right
- GAN research can stop layering empirical tricks: a clean, modern backbone trained with a stable loss outperforms the best-known trick-laden GAN on the tested benchmarks.
- One-step adversarial generators can match or beat multi-step diffusion models on FID without distillation, so the computational advantage of GANs is achievable in practice.
- Loss design becomes a checkable engineering step: candidate relativistic losses can be screened for local convergence by the same eigenvalue test used here.
- The paper's mode-recovery numbers on Stacked-MNIST indicate that penalizing the discriminator on both real and fake data (R1 and R2 together) is the practical route to global stability and full mode coverage, even though the local theorem holds for either penalty alone.
- Because the reported FID gains are obtained without a pretrained ImageNet discriminator, the results are not attributable to feature leakage the way some state-of-the-art GANs' numbers are.
Reading between the lines
- A natural extension we draw, not the paper's: the same recipe of a stable adversarial loss plus a modern backbone could be carried into text-to-image generation, where one-step generators would compete with diffusion without needing consistency distillation.
- Because removing the style network also removes the standard interface for editing and inversion, we expect R3GAN-type generators to need new mechanisms for controllability if they become the common baseline.
- We would predict, based on the DiracRpGAN analysis, that other relativistic losses whose unregularized Jacobians have imaginary-axis eigenvalues will also exhibit non-convergence unless equipped with analogous zero-centered penalties; this is a testable classification the paper does not perform.
- A cheaper empirical check of our own: monitoring $\mathbb{E}_{x\sim p_\theta}\|\nabla_x D\|^2$ during R1-only training should show the fake-side gradient norm exploding before the generator loss diverges, directly testing the paper's proposed failure mechanism.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper argues that GAN training need not rely on a large bag of empirical tricks. It derives a regularized relativistic GAN objective (RpGAN + R1 + R2), proves local convergence for the R1- or R2-regularized variants in the idealized setting of Mescheder et al., and then uses this loss to strip StyleGAN2 down to a minimal backbone and modernize it into R3GAN. The authors report FID results on FFHQ-256, FFHQ-64, CIFAR-10, ImageNet-32, ImageNet-64, and StackedMNIST, claiming that R3GAN surpasses StyleGAN2 and is competitive with state-of-the-art GANs and diffusion models.
Significance. If the empirical results are reproducible, this is a valuable contribution: it provides a simple modern GAN baseline, a careful ablation roadmap from StyleGAN2 to R3GAN, and evidence that a principled loss can replace many ad-hoc GAN tricks. The paper is commendably transparent in several respects: it reports detailed hyperparameters (Appendix D), includes training curves (Appendix G), documents negative results (Appendix E), and adapts the Mescheder et al. convergence framework to RpGAN. However, the gap between the theoretical guarantee and the actual training recipe, the unsupported ImageNet claim in the abstract, and the absence of error bars currently prevent me from endorsing the paper as is.
major comments (3)
- [Abstract; Sections 4.6-4.7] The abstract claims that R3GAN 'surpasses StyleGAN2 on FFHQ, ImageNet, CIFAR, and Stacked MNIST datasets,' but Tables 7 and 8 (ImageNet-32 and ImageNet-64) contain no StyleGAN2 baseline; the comparisons there are against StyleGAN-XL, BigGAN-deep, and diffusion models. The ImageNet part of the claim is therefore unsupported by the reported experiments. Please add the missing StyleGAN2 results or revise the claim to name the methods actually compared.
- [Section 2.4; Appendix C, Eq. (20)] The convergence theorem is proved for RpGAN with either R1 or R2 alone, yet the paper's own Figure 1 and Table 1 show that RpGAN+R1 diverges from random initialization on StackedMNIST. The statement in Appendix C that the combined R1+R2 case follows 'trivially' is too thin, because the final recipe relies on exactly this combination, and the practical model also uses augmentation, EMA, cosine schedules, and BFloat16 (Appendix D, Table 9) that are absent from the analysis. The load-bearing claim that the loss is 'well-behaved' and removes the need for tricks is thus not established by the proof; it is an empirical property of a specific architecture and training schedule. Please provide the combined proof or a detailed argument, and explicitly scope the theoretical claim as local and as not covering the full training recipe.
- [Tables 4-8; NeurIPS checklist item 7] All FID numbers are single runs without error bars or confidence intervals. Some of the headline improvements are modest (e.g., 2.75 vs. 3.78 on FFHQ-256, 1.95 vs. 2.39 on FFHQ-64, 1.96 vs. 2.42 on CIFAR-10), so run-to-run variation could affect the conclusions. Please provide multiple seeds and variance estimates for at least the smaller-scale experiments (StackedMNIST, CIFAR-10, FFHQ-64), and state the known variability for the larger runs.
minor comments (4)
- [Eq. (20)] R2 is written as R2(θ,ψ), but the gradient vector field only includes −∇ψRi and not a generator-side gradient ∇θR2; please clarify that R2 is a discriminator-only regularizer in the training algorithm.
- [Table 2 caption] There is a typo: 'evaluted' should be 'evaluated'.
- [Section 4.2] The text says the comparison includes 'likelihood-based methods,' but Table 3 also lists GAN variants such as MEG and Adv. DSM; please clarify the category labels for these baselines.
- [NeurIPS checklist item 5] The checklist states that no code is available at submission time; given that the paper proposes a new baseline architecture and loss, please make the code publicly available or provide a detailed implementation plan at revision time.
Circularity Check
No significant circularity found: the loss derivation and convergence analysis build on external prior work, and the empirical results are genuine outputs rather than fitted inputs.
full rationale
The paper's central derivation chain is self-contained in the relevant sense. The proposed loss is RpGAN (Jolicoeur-Martineau, external) combined with R1/R2 zero-centered gradient penalties (Mescheder et al. and Roth et al., external). The local convergence theorem in Appendix C adapts Mescheder et al.'s proof technique to the relativistic pairing loss, with modified assumptions stated explicitly; the key spectral lemmas are cited to Mescheder et al. rather than assumed from the paper's own conclusions. The DiracRpGAN non-convergence proof is a concrete counterexample and not a restatement of the desired stability claim. The claim that R1 alone is not sufficient for global training stability is explicitly presented as an empirical demonstration, and the paper acknowledges that the theory only covers local convergence. The FID and mode-coverage results are measured outcomes of training runs, not parameters fitted to those benchmarks. The self-citations present in the reference list are unrelated to the load-bearing loss derivation or convergence proof. The most substantive weakness, that the theorem covers R1 or R2 separately while the practical recipe uses both with the combined case asserted as 'trivially extended,' is a gap between theory and practice, not circularity. Similarly, the use of augmentation, EMA, cosine schedules, and BFloat16 despite the claim of discarding ad-hoc tricks is an overclaim about scope, not a definitional or self-citational reduction. Overall, no step in the claimed derivation reduces by construction to its own inputs.
Assumptions & free parameters
free parameters (6)
- R1/R2 regularization coefficient gamma =
per-dataset schedules, e.g. 150 to 15 (FFHQ-256), 0.05 to 0.005 (CIFAR-10)
- Learning rate schedule =
2e-4 to 5e-5 for CIFAR-10, FFHQ, and ImageNet; 2e-4 constant for StackedMNIST
- Adam beta2 schedule =
0.9 to 0.99
- EMA half-life schedule =
e.g. 0 to 5 Mimg (CIFAR-10), 0 to 50 Mimg (ImageNet)
- Augmentation probability schedule =
0 to 0.55 (CIFAR-10), 0 to 0.3 (FFHQ), 0 to 0.5 (ImageNet-32)
- Network width, channels per resolution, and group sizes =
e.g. 768-768-768-768 with groups 96 (CIFAR-10); 96-192-384-768-768-768-768 with groups 12-24-48-96-96-96-96 (FFHQ)
assumptions (5)
- domain assumption Assumption I: p_theta* = p_D and D_psi* is constant on a neighborhood of supp p_D (realizable case).
- standard math Assumption II: f'(0) != 0 and f''(0) < 0 for the loss function f.
- domain assumption Assumption III: equilibrium manifolds M_G and M_D are C1, and the discriminator can detect any deviation from generator equilibrium.
- standard math Mescheder et al.'s local convergence framework: eigenvalues of the Jacobian of the gradient vector field determine convergence for small learning rates.
- domain assumption Sun et al.'s landscape theorem: the RpGAN objective has no mode-dropping bad local minima and every basin is global.
Cite this review
Pith. "Pith review of The GAN is dead; long live the GAN! A Modern GAN Baseline." pith.science (2026). https://pith.science/paper/5JU3BFBT
@misc{pith2026250105441,
author = {Pith},
title = {Pith review of: The GAN is dead; long live the GAN! A Modern GAN Baseline},
year = {2026},
howpublished = {\url{https://pith.science/paper/5JU3BFBT}},
note = {Machine review of arXiv:2501.05441}
}
read the original abstract
There is a widely-spread claim that GANs are difficult to train, and GAN architectures in the literature are littered with empirical tricks. We provide evidence against this claim and build a modern GAN baseline in a more principled manner. First, we derive a well-behaved regularized relativistic GAN loss that addresses issues of mode dropping and non-convergence that were previously tackled via a bag of ad-hoc tricks. We analyze our loss mathematically and prove that it admits local convergence guarantees, unlike most existing relativistic losses. Second, our new loss allows us to discard all ad-hoc tricks and replace outdated backbones used in common GANs with modern architectures. Using StyleGAN2 as an example, we present a roadmap of simplification and modernization that results in a new minimalist baseline -- R3GAN. Despite being simple, our approach surpasses StyleGAN2 on FFHQ, ImageNet, CIFAR, and Stacked MNIST datasets, and compares favorably against state-of-the-art GANs and diffusion models.
Figures
Figures from the paper (12 more)
Reference graph
Works this paper leans on
-
[1]
Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. Layer normalization. arXiv preprint arXiv:1607.06450, 2016
arXiv 2016
-
[2]
Smu: smooth activation function for deep networks using smoothing maximum technique
Koushik Biswas, Sandeep Kumar, Shilpak Banerjee, and Ashish Kumar Pandey. Smu: smooth activation function for deep networks using smoothing maximum technique. arXiv preprint arXiv:2111.04682, 2021
arXiv 2021
-
[3]
Large scale gan training for high fidelity natural image synthesis
Andrew Brock, Jeff Donahue, and Karen Simonyan. Large scale gan training for high fidelity natural image synthesis. arXiv preprint arXiv:1809.11096, 2018
arXiv 2018
-
[4]
High-performance large-scale image recognition without normalization
Andy Brock, Soham De, Samuel L Smith, and Karen Simonyan. High-performance large-scale image recognition without normalization. In International Conference on Machine Learning, pp. 1059–1071. PMLR, 2021
2021
-
[5]
Xception: Deep learning with depthwise separable convolutions
François Chollet. Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1251–1258, 2017
2017
-
[6]
A downsampled variant of imagenet as an alternative to the cifar datasets
Patryk Chrabaszcz, Ilya Loshchilov, and Frank Hutter. A downsampled variant of imagenet as an alternative to the cifar datasets. arXiv preprint arXiv:1707.08819, 2017
arXiv 2017
-
[7]
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
2021
-
[8]
Prescribed generative adversarial networks
Adji B Dieng, Francisco JR Ruiz, David M Blei, and Michalis K Titsias. Prescribed generative adversarial networks. arXiv preprint arXiv:1910.04302, 2019
arXiv 1910
Show all 114 references
-
[9]
An image is worth 16x16 words: Transformers for image recognition at scale
Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv...
2010 arXiv
-
[10]
DigGAN: Discriminator gradient gap regularization for GAN training with limited data
Tiantian Fang, Ruoyu Sun, and Alex Schwing. DigGAN: Discriminator gradient gap regularization for GAN training with limited data. In Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho (eds.), Advances in Neural Information Processing Systems, 2022. URL https://op...
2022
-
[11]
Negative momentum for improved game dynamics
Gauthier Gidel, Reyhane Askari Hemmat, Mohammad Pezeshki, Rémi Le Priol, Gabriel Huang, Simon Lacoste-Julien, and Ioannis Mitliagkas. Negative momentum for improved game dynamics. In The 22nd International Conference on Artificial Intelligence and Statistics, pp. 1802–1811. PMLR, 2019
2019
-
[12]
Commoncanvas: Open diffusion models trained on creative-commons images
Aaron Gokaslan, A Feder Cooper, Jasmine Collins, Landan Seguin, Austin Jacobson, Mihir Patel, Jonathan Frankle, Cory Stephenson, and V olodymyr Kuleshov. Commoncanvas: Open diffusion models trained on creative-commons images. In Proceedings of the IEEE/CVF Conference on Comput...
2024
-
[13]
Generative adversarial networks
Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks. Communications of the ACM, 63(11): 139–144, 2020
2020
-
[14]
Improved training of wasserstein gans
Ishaan Gulrajani, Faruk Ahmed, Martin Arjovsky, Vincent Dumoulin, and Aaron C Courville. Improved training of wasserstein gans. Advances in neural information processing systems, 30, 2017
2017
-
[15]
Delving deep into rectifiers: Surpassing human-level performance on imagenet classification
Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. In Proceedings of the IEEE international conference on computer vision, pp. 1026–1034, 2015
2015
-
[16]
Deep residual learning for image recognition
Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778, 2016. 11
2016
-
[17]
Identity mappings in deep residual networks
Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Identity mappings in deep residual networks. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part IV 14, pp. 630–645. Springer, 2016
2016
-
[18]
Gaussian error linear units (gelus)
Dan Hendrycks and Kevin Gimpel. Gaussian error linear units (gelus). arXiv preprint arXiv:1606.08415, 2016
2016 arXiv
-
[19]
Gans trained by a two time-scale update rule converge to a local nash equilibrium
Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems, 30, 2017
2017
-
[20]
Denoising diffusion probabilistic models
Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020
2020
-
[21]
Batch normalization: Accelerating deep network training by reducing internal covariate shift
Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International conference on machine learning, pp. 448–456. pmlr, 2015
2015
-
[22]
The relativistic discriminator: a key element missing from standard gan
Alexia Jolicoeur-Martineau. The relativistic discriminator: a key element missing from standard gan. arXiv preprint arXiv:1807.00734, 2018
2018 arXiv
-
[23]
Gradient penalty from a maximum margin perspective
Alexia Jolicoeur-Martineau and Ioannis Mitliagkas. Gradient penalty from a maximum margin perspective. arXiv preprint arXiv:1910.06922, 2019
1910 arXiv
-
[24]
Ad- versarial score matching and improved sampling for image generation
Alexia Jolicoeur-Martineau, Rémi Piché-Taillefer, Rémi Tachet des Combes, and Ioannis Mitliagkas. Ad- versarial score matching and improved sampling for image generation. arXiv preprint arXiv:2009.05475, 2020
2009 arXiv
-
[25]
Studiogan: a taxonomy and benchmark of gans for image synthesis
Minguk Kang, Joonghyuk Shin, and Jaesik Park. Studiogan: a taxonomy and benchmark of gans for image synthesis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023
2023
-
[26]
Scaling up gans for text-to-image synthesis
Minguk Kang, Jun-Yan Zhu, Richard Zhang, Jaesik Park, Eli Shechtman, Sylvain Paris, and Taesung Park. Scaling up gans for text-to-image synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10124–10134, 2023
2023
-
[27]
Msg-gan: Multi-scale gradients for generative adversarial networks
Animesh Karnewar and Oliver Wang. Msg-gan: Multi-scale gradients for generative adversarial networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 7799–7808, 2020
2020
-
[28]
Progressive growing of gans for improved quality, stability, and variation
Tero Karras, Timo Aila, Samuli Laine, and Jaakko Lehtinen. Progressive growing of gans for improved quality, stability, and variation. arXiv preprint arXiv:1710.10196, 2017
2017 arXiv
-
[29]
A style-based generator architecture for generative adversarial networks
Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4401–4410, 2019
2019
-
[30]
Training generative adversarial networks with limited data
Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, and Timo Aila. Training generative adversarial networks with limited data. Advances in neural information processing systems, 33:12104–12114, 2020
2020
-
[31]
Analyzing and improving the image quality of stylegan
Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improving the image quality of stylegan. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 8110–8119, 2020
2020
-
[32]
Alias-free generative adversarial networks
Tero Karras, Miika Aittala, Samuli Laine, Erik Härkönen, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Alias-free generative adversarial networks. Advances in Neural Information Processing Systems, 34:852–863, 2021
2021
-
[33]
Elucidating the design space of diffusion-based generative models
Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models. Advances in Neural Information Processing Systems, 35:26565–26577, 2022
2022
-
[34]
Analyzing and improving the training dynamics of diffusion models
Tero Karras, Miika Aittala, Jaakko Lehtinen, Janne Hellsten, Timo Aila, and Samuli Laine. Analyzing and improving the training dynamics of diffusion models. arXiv preprint arXiv:2312.02696, 2023
2023 arXiv
-
[35]
Soft truncation: A universal training technique of score-based diffusion model for high precision score estimation
Dongjun Kim, Seungjae Shin, Kyungwoo Song, Wanmo Kang, and Il-Chul Moon. Soft truncation: A universal training technique of score-based diffusion model for high precision score estimation. arXiv preprint arXiv:2106.05527, 2021
2021 arXiv
-
[36]
Variational diffusion models
Diederik Kingma, Tim Salimans, Ben Poole, and Jonathan Ho. Variational diffusion models. Advances in neural information processing systems, 34:21696–21707, 2021. 12
2021
-
[37]
Learning multiple layers of features from tiny images
Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. Thesis, 2009
2009
-
[38]
Imagenet classification with deep convolutional neural networks
Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. In F. Pereira, C.J. Burges, L. Bottou, and K.Q. Wein- berger (eds.), Advances in Neural Information Processing Systems , volume 25. Curran Asso- ciates, Inc...
2012
-
[39]
Maximum entropy generators for energy-based models
Rithesh Kumar, Sherjil Ozair, Anirudh Goyal, Aaron Courville, and Yoshua Bengio. Maximum entropy generators for energy-based models. arXiv preprint arXiv:1901.08508, 2019
1901 arXiv
-
[40]
Improved precision and recall metric for assessing generative models
Tuomas Kynkäänniemi, Tero Karras, Samuli Laine, Jaakko Lehtinen, and Timo Aila. Improved precision and recall metric for assessing generative models. Advances in neural information processing systems, 32, 2019
2019
-
[41]
The role of imagenet classes in fréchet inception distance
Tuomas Kynkäänniemi, Tero Karras, Miika Aittala, Timo Aila, and Jaakko Lehtinen. The role of imagenet classes in fréchet inception distance. arXiv preprint arXiv:2203.06026, 2022
2022 arXiv
-
[42]
Vitgan: Training gans with vision transformers
Kwonjoon Lee, Huiwen Chang, Lu Jiang, Han Zhang, Zhuowen Tu, and Ce Liu. Vitgan: Training gans with vision transformers. arXiv preprint arXiv:2107.04589, 2021
2021 arXiv
-
[43]
Enhanced deep residual networks for single image super-resolution
Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. Enhanced deep residual networks for single image super-resolution. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pp. 136–144, 2017
2017
-
[44]
Geometric gan
Jae Hyun Lim and Jong Chul Ye. Geometric gan. arXiv preprint arXiv:1705.02894, 2017
2017 arXiv
-
[45]
Anycost gans for interactive image synthesis and editing
Ji Lin, Richard Zhang, Frieder Ganz, Song Han, and Jun-Yan Zhu. Anycost gans for interactive image synthesis and editing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 14986–14996, 2021
2021
-
[46]
Pacgan: The power of two samples in generative adversarial networks
Zinan Lin, Ashish Khetan, Giulia Fanti, and Sewoong Oh. Pacgan: The power of two samples in generative adversarial networks. Advances in neural information processing systems, 31, 2018
2018
-
[47]
Swin transformer: Hierarchical vision transformer using shifted windows
Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pp. 10012–10022, 2021
2021
-
[48]
A convnet for the 2020s
Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, and Saining Xie. A convnet for the 2020s. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11976–11986, 2022
2022
-
[49]
Compensation sampling for improved convergence in diffusion models
Hui Lu, Ronald Poppe, et al. Compensation sampling for improved convergence in diffusion models. arXiv preprint arXiv:2312.06285, 2023
2023 arXiv
-
[50]
Least squares generative adversarial networks
Xudong Mao, Qing Li, Haoran Xie, Raymond YK Lau, Zhen Wang, and Stephen Paul Smolley. Least squares generative adversarial networks. InProceedings of the IEEE international conference on computer vision, pp. 2794–2802, 2017
2017
-
[51]
The numerics of gans
Lars Mescheder, Sebastian Nowozin, and Andreas Geiger. The numerics of gans. Advances in neural information processing systems, 30, 2017
2017
-
[52]
Which training methods for gans do actually converge? In International conference on machine learning, pp
Lars Mescheder, Andreas Geiger, and Sebastian Nowozin. Which training methods for gans do actually converge? In International conference on machine learning, pp. 3481–3490. PMLR, 2018
2018
-
[53]
Unrolled generative adversarial networks
Luke Metz, Ben Poole, David Pfau, and Jascha Sohl-Dickstein. Unrolled generative adversarial networks. In International Conference on Learning Representations, 2016
2016
-
[54]
cgans with projection discriminator
Takeru Miyato and Masanori Koyama. cgans with projection discriminator. arXiv preprint arXiv:1802.05637, 2018
2018 arXiv
-
[55]
Gradient descent gan optimization is locally stable
Vaishnavh Nagarajan and J Zico Kolter. Gradient descent gan optimization is locally stable. Advances in neural information processing systems, 30, 2017
2017
-
[56]
Input perturbation reduces exposure bias in diffusion models
Mang Ning, Enver Sangineto, Angelo Porrello, Simone Calderara, and Rita Cucchiara. Input perturbation reduces exposure bias in diffusion models. arXiv preprint arXiv:2301.11706, 2023
2023 arXiv
-
[57]
f-gan: Training generative neural samplers using variational divergence minimization
Sebastian Nowozin, Botond Cseke, and Ryota Tomioka. f-gan: Training generative neural samplers using variational divergence minimization. Advances in neural information processing systems, 29, 2016. 13
2016
-
[58]
Scalable diffusion models with transformers
William Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 4195–4205, 2023
2023
-
[59]
Diffusion autoencoders: Toward a meaningful and decodable representation
Konpat Preechakul, Nattanat Chatthee, Suttisak Wizadwongsa, and Supasorn Suwajanakorn. Diffusion autoencoders: Toward a meaningful and decodable representation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 10619–10629, June 2022
2022
-
[60]
Unsupervised representation learning with deep convolutional generative adversarial networks
Alec Radford, Luke Metz, and Soumith Chintala. Unsupervised representation learning with deep convolutional generative adversarial networks. arXiv preprint arXiv:1511.06434, 2015
2015 arXiv
-
[61]
Searching for activation functions
Prajit Ramachandran, Barret Zoph, and Quoc V Le. Searching for activation functions. arXiv preprint arXiv:1710.05941, 2017
2017 arXiv
-
[62]
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, pp. 10684–10695, 2022
2022
-
[63]
U-net: Convolutional networks for biomedical image segmentation
Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part ...
2015
-
[64]
Stabilizing training of generative adversarial networks through regularization
Kevin Roth, Aurelien Lucchi, Sebastian Nowozin, and Thomas Hofmann. Stabilizing training of generative adversarial networks through regularization. Advances in neural information processing systems, 30, 2017
2017
-
[65]
Lite- vae: Lightweight and efficient variational autoencoders for latent diffusion models
Seyedmorteza Sadat, Jakob Buhmann, Derek Bradley, Otmar Hilliges, and Romann M Weber. Lite- vae: Lightweight and efficient variational autoencoders for latent diffusion models. arXiv preprint arXiv:2405.14477, 2024
2024 arXiv
-
[66]
Diffusion models with learned adaptive noise
Subham Sekhar Sahoo, Aaron Gokaslan, Chris De Sa, and V olodymyr Kuleshov. Diffusion models with learned adaptive noise. arXiv preprint arXiv:2312.13236, 2023
2023 arXiv
-
[67]
Mobilenetv2: Inverted residuals and linear bottlenecks
Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 4510–4520, 2018
2018
-
[68]
Projected gans converge faster
Axel Sauer, Kashyap Chitta, Jens Müller, and Andreas Geiger. Projected gans converge faster. Advances in Neural Information Processing Systems, 34:17480–17492, 2021
2021
-
[69]
StyleGAN-XL: Scaling stylegan to large diverse datasets
Axel Sauer, Katja Schwarz, and Andreas Geiger. StyleGAN-XL: Scaling stylegan to large diverse datasets. In ACM SIGGRAPH 2022 conference proceedings, pp. 1–10, 2022
2022
-
[70]
Stylegan-t: Unlocking the power of gans for fast large-scale text-to-image synthesis
Axel Sauer, Tero Karras, Samuli Laine, Andreas Geiger, and Timo Aila. Stylegan-t: Unlocking the power of gans for fast large-scale text-to-image synthesis. In International conference on machine learning, pp. 30105–30118. PMLR, 2023
2023
-
[72]
Real-time single image and video super-resolution using an efficient sub- pixel convolutional neural network
Wenzhe Shi, Jose Caballero, Ferenc Huszár, Johannes Totz, Andrew P Aitken, Rob Bishop, Daniel Rueckert, and Zehan Wang. Real-time single image and video super-resolution using an efficient sub- pixel convolutional neural network. In Proceedings of the IEEE conference on comput...
2016
-
[73]
Very deep convolutional networks for large-scale image recognition
Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014
2014 arXiv
-
[74]
Polynomial implicit neural representations for large diverse datasets
Rajhans Singh, Ankita Shukla, and Pavan Turaga. Polynomial implicit neural representations for large diverse datasets. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 2041–2051, 2023
2023
-
[75]
Amortised map inference for image super-resolution
Casper Kaae Sønderby, Jose Caballero, Lucas Theis, Wenzhe Shi, and Ferenc Huszár. Amortised map inference for image super-resolution. arXiv preprint arXiv:1610.04490, 2016
2016 arXiv
-
[76]
Denoising diffusion implicit models
Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. In International Conference on Learning Representations, 2021. 14
2021
-
[77]
Improved techniques for training consistency models
Yang Song and Prafulla Dhariwal. Improved techniques for training consistency models. In The Twelfth International Conference on Learning Representations, 2024
2024
-
[78]
Score-based generative modeling through stochastic differential equations
Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020
2011 arXiv
-
[79]
Consistency models
Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. Consistency models. In International Conference on Machine Learning, pp. 32211–32252. PMLR, 2023
2023
-
[80]
Veegan: Reducing mode collapse in gans using implicit variational learning
Akash Srivastava, Lazar Valkov, Chris Russell, Michael U Gutmann, and Charles Sutton. Veegan: Reducing mode collapse in gans using implicit variational learning. Advances in neural information processing systems, 30, 2017
2017
-
[81]
Towards a better global loss landscape of gans
Ruoyu Sun, Tiantian Fang, and Alexander Schwing. Towards a better global loss landscape of gans. Advances in Neural Information Processing Systems, 33:10186–10198, 2020
2020
-
[82]
SAN: Inducing metrizability of GAN with discriminative normalized linear layer
Yuhta Takida, Masaaki Imaizumi, Takashi Shibuya, Chieh-Hsin Lai, Toshimitsu Uesaka, Naoki Murata, and Yuki Mitsufuji. SAN: Inducing metrizability of GAN with discriminative normalized linear layer. In The Twelfth International Conference on Learning Representations, 2024. URL ...
2024
-
[83]
Alleviation of gradient exploding in gans: Fake can be real
Song Tao and Jia Wang. Alleviation of gradient exploding in gans: Fake can be real. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 1191–1200, 2020
2020
-
[84]
Improving generalization and stability of generative adversarial networks
Hoang Thanh-Tung, Truyen Tran, and Svetha Venkatesh. Improving generalization and stability of generative adversarial networks. In International Conference on Learning Representations, 2019. URL https://openreview.net/forum?id=ByxPYjC5KQ
2019
-
[85]
Instance normalization: The missing ingredient for fast stylization
Dmitry Ulyanov, Andrea Vedaldi, and Victor Lempitsky. Instance normalization: The missing ingredient for fast stylization. arXiv preprint arXiv:1607.08022, 2016
2016 arXiv
-
[86]
Score-based generative modeling in latent space
Arash Vahdat, Karsten Kreis, and Jan Kautz. Score-based generative modeling in latent space. Advances in neural information processing systems, 34:11287–11302, 2021
2021
-
[87]
Attention is all you need
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in neural information processing systems, 30, 2017
2017
-
[88]
Esrgan: Enhanced super-resolution generative adversarial networks
Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy. Esrgan: Enhanced super-resolution generative adversarial networks. In Proceedings of the European conference on computer vision (ECCV) workshops, pp. 0–0, 2018
2018
-
[89]
Infodiffusion: Representation learning using information maximizing diffusion models
Yingheng Wang, Yair Schiff, Aaron Gokaslan, Weishen Pan, Fei Wang, Christopher De Sa, and V olodymyr Kuleshov. Infodiffusion: Representation learning using information maximizing diffusion models. In International Conference on Machine Learning, pp. 36336–36354. PMLR, 2023
2023
-
[90]
Diffusion-gan: Training gans with diffusion
Zhendong Wang, Huangjie Zheng, Pengcheng He, Weizhu Chen, and Mingyuan Zhou. Diffusion-gan: Training gans with diffusion. In The Eleventh International Conference on Learning Representations, 2023
2023
-
[92]
Group normalization
Yuxin Wu and Kaiming He. Group normalization. In Proceedings of the European conference on computer vision (ECCV), pp. 3–19, 2018
2018
-
[93]
Vaebm: A symbiosis between variational autoencoders and energy-based models
Zhisheng Xiao, Karsten Kreis, Jan Kautz, and Arash Vahdat. Vaebm: A symbiosis between variational autoencoders and energy-based models. arXiv preprint arXiv:2010.00654, 2020
2010 arXiv
-
[94]
Tackling the generative learning trilemma with denoising diffusion gans
Zhisheng Xiao, Karsten Kreis, and Arash Vahdat. Tackling the generative learning trilemma with denoising diffusion gans. arXiv preprint arXiv:2112.07804, 2021
2021 arXiv
-
[95]
Aggregated residual transforma- tions for deep neural networks
Saining Xie, Ross Girshick, Piotr Dollár, Zhuowen Tu, and Kaiming He. Aggregated residual transforma- tions for deep neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1492–1500, 2017
2017
-
[96]
One-step diffusion with distribution matching distillation
Tianwei Yin, Michaël Gharbi, Richard Zhang, Eli Shechtman, Fredo Durand, William T Freeman, and Taesung Park. One-step diffusion with distribution matching distillation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 6613–6623, 2024. 15
2024
-
[97]
Metaformer is actually what you need for vision
Weihao Yu, Mi Luo, Pan Zhou, Chenyang Si, Yichen Zhou, Xinchao Wang, Jiashi Feng, and Shuicheng Yan. Metaformer is actually what you need for vision. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 10819–10829, 2022
2022
-
[98]
Styleswin: Transformer-based gan for high-resolution image generation
Bowen Zhang, Shuyang Gu, Bo Zhang, Jianmin Bao, Dong Chen, Fang Wen, Yong Wang, and Baining Guo. Styleswin: Transformer-based gan for high-resolution image generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 11304–11314, 2022
2022
-
[99]
Fixup initialization: Residual learning without normalization
Hongyi Zhang, Yann N Dauphin, and Tengyu Ma. Fixup initialization: Residual learning without normalization. arXiv preprint arXiv:1901.09321, 2019
1901 arXiv
-
[100]
Making convolutional networks shift-invariant again
Richard Zhang. Making convolutional networks shift-invariant again. In International conference on machine learning, pp. 7324–7334. PMLR, 2019
2019
-
[101]
Improved consistency regularization for gans
Zhengli Zhao, Sameer Singh, Honglak Lee, Zizhao Zhang, Augustus Odena, and Han Zhang. Improved consistency regularization for gans. In Proceedings of the AAAI conference on artificial intelligence , volume 35, pp. 11033–11041, 2021. 16 Appendices A Local convergence Following ...
2021
-
[102]
Claim of convergence properties is justified in Appendices A,B,C
Claims Question: Do the main claims made in the abstract and introduction accurately reflect the paper’s contributions and scope? Answer: [Yes] Justification: Claim of stability is justified by Figure 1 and later experimental performance. Claim of convergence properties is jus...
-
[103]
Limitations
Limitations Question: Does the paper discuss the limitations of the work performed by the authors? Answer: [Yes] Justification: Please see Section 5. Guidelines: • The answer NA means that the paper has no limitation while the answer No means that the paper has limitations, bu...
-
[104]
[52] is required, but this is cited appropri- ately to help the reader
Theory Assumptions and Proofs Question: For each theoretical result, does the paper provide the full set of assumptions and a complete (and correct) proof? 33 Answer: [Yes] Justification: Prior knowledge of Mescheder et al. [52] is required, but this is cited appropri- ately t...
-
[105]
Guidelines: • The answer NA means that the paper does not include experiments
Experimental Result Reproducibility Question: Does the paper fully disclose all the information needed to reproduce the main ex- perimental results of the paper to the extent that it affects the main claims and/or conclusions of the paper (regardless of whether the code and da...
-
[106]
There is no code at submission time
Open access to data and code Question: Does the paper provide open access to the data and code, with sufficient instruc- tions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [No] Justification: There is no new data. There ...
-
[107]
Guidelines: • The answer NA means that the paper does not include experiments
Experimental Setting/Details Question: Does the paper specify all the training and test details (e.g., data splits, hyper- parameters, how they were chosen, type of optimizer, etc.) necessary to understand the results? Answer: [Yes] Justification: Supplemental table lists all ...
-
[108]
We do not have the compute time to provide variance bars on training executions
Experiment Statistical Significance Question: Does the paper report error bars suitably and correctly defined or other appropriate information about the statistical significance of the experiments? Answer: [No] Justification: Each experiment takes many days to compute, some ta...
-
[109]
Guidelines: • The answer NA means that the paper does not include experiments
Experiments Compute Resources Question: For each experiment, does the paper provide sufficient information on the com- puter resources (type of compute workers, memory, time of execution) needed to reproduce the experiments? Answer: [Yes] Justification: Please see supplemental...
-
[110]
Guidelines: • The answer NA means that the authors have not reviewed the NeurIPS Code of Ethics
Code Of Ethics Question: Does the research conducted in the paper conform, in every respect, with the NeurIPS Code of Ethics https://neurips.cc/public/EthicsGuidelines? Answer: [Yes] Justification: Experimental settings are standard and within the norms of the community. Guide...
-
[111]
The paper describes a basic machine learning methodology, and so does not address a specific application with specific societal impacts
Broader Impacts Question: Does the paper discuss both potential positive societal impacts and negative societal impacts of the work performed? Answer: [Yes] Justification: We mention it briefly in Section 5. The paper describes a basic machine learning methodology, and so does...
-
[112]
The cost of training these large GANs is not prohibitive and is often done by hobbyists
Safeguards Question: Does the paper describe safeguards that have been put in place for responsible release of data or models that have a high risk for misuse (e.g., pretrained language models, image generators, or scraped datasets)? Answer: [No] Justification: There is no new...
-
[113]
Guidelines: • The answer NA means that the paper does not use existing assets
Licenses for existing assets Question: Are the creators or original owners of assets (e.g., code, data, models), used in the paper, properly credited and are the license and terms of use explicitly mentioned and properly respected? Answer: [Yes] Justification: All datasets are...
-
[114]
Guidelines: • The answer NA means that the paper does not release new assets
New Assets Question: Are new assets introduced in the paper well documented and is the documentation provided alongside the assets? Answer: [NA] Justification: No new assets are released. Guidelines: • The answer NA means that the paper does not release new assets. • Researche...
-
[115]
Guidelines: • The answer NA means that the paper does not involve crowdsourcing nor research with human subjects
Crowdsourcing and Research with Human Subjects Question: For crowdsourcing experiments and research with human subjects, does the paper include the full text of instructions given to participants and screenshots, if applicable, as well as details about compensation (if any)? A...
-
[116]
Guidelines: 38 • The answer NA means that the paper does not involve crowdsourcing nor research with human subjects
Institutional Review Board (IRB) Approvals or Equivalent for Research with Human Subjects Question: Does the paper describe potential risks incurred by study participants, whether such risks were disclosed to the subjects, and whether Institutional Review Board (IRB) approvals...
Reviewed August 10, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.