Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

A solvable generative model with a linear, one-step denoiser

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A solvable diffusion model yields an exact formula for the generation gap.

desk verdict The paper's central theorem is proved for a different regression model than the one defined in Section 2, leaving the main formula unsupported; still, the setup is a good idea and the flaw is fixable. read the letter →

arxiv 2411.17807 v3 pith:6UMFUXUF submitted 2024-11-26 cs.LG cs.CV

Indranil Halder This is my paper · ORCID
classification cs.LGcs.CV MSC 60B2062J0768T07
keywords diffusionmodelslineardenoiserKullback-Leiblerdivergencehigh-dimensionalasymptoticsdeterministicequivalencerandommatrixtheorygenerationgapsamplecomplexity
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

This paper constructs a deliberately simple generative model—a single linear denoising step trained on finitely many samples—and asks exactly how far its output distribution is from the true distribution it was trained on. The main result is an explicit formula for the expected Kullback-Leibler divergence between generated and sampling distributions for isotropic Gaussian data, valid when the dimension $d$ and training size $n$ grow together and the noise scale is small. The formula shows that the divergence starts falling monotonically with more data once $n$ exceeds $d$, a much friendlier threshold than the exponential scale suggested by the curse of dimensionality. The same calculation is then used to explain a practical observation: in realistic diffusion models, increasing the number of diffusion steps improves output quality.

What carries the argument

The central object is a linear denoiser $\hat Y = \hat\theta_0 + \hat\theta_1 X$ trained by least squares, paired with the KL divergence decomposed into mean and variance parts, $\mathrm{KL} = \mathrm{KL}_{\mathrm{mean}} + \mathrm{KL}_{\mathrm{var}}$. The variance part is computed in the $n,d\to\infty$ limit using the principle of deterministic equivalence from free random-matrix theory: the empirical covariance of the design matrix is treated as a free product of the population covariance with a white Wishart matrix, giving explicit traces $\mathrm{Tr}\,\hat\Sigma_{\theta_1}$ and $\mathrm{Tr}\,\hat\Sigma_{\theta_1}^2$ via Stieltjes-transform identities. This carries the argument because it turns the random least-squares estimator into a deterministic function of $\alpha=d/n$, the diffusion time $T$, and the noise parameter $\hat\lambda$.

What would settle it

Simulate the Section 2 model at $d=2000$, $n=4000$ ($\alpha=0.5$), $T=2$, and $\hat\lambda=10^{-3}$, and compare the seed-averaged empirical $\mathrm{KL}_{\mathrm{var}}$ with the $\alpha<1$ formula (2.7); agreement to first order in $\hat\lambda$ would confirm the independence assumption, while a mismatch linear in $\hat\lambda$ would show the errors-in-variables correlation matters.

Watch

Extended reading notes

Core claim

In the high-dimensional limit $n,d\to\infty$ with $\alpha=d/n$ fixed and noise scale $\lambda=\hat\lambda\sigma^2 e^{-2T}$ small, the expected variance contribution to the KL divergence between the generated Gaussian and the true isotropic Gaussian sampling distribution is given by closed-form expressions. For $\alpha<1$ the paper derives $$\langle \mathrm{KL}_{\mathrm{var}}\rangle = \frac{d\$\alpha$\hat\$\lambda$ $e^{{-4T}}$($e^{{2T}}$-1)}{2(1-\$\alpha$)} + \frac{d\hat\$lambda^{2}$ $e^{{-8T}}$($e^{{2T}}$-1)^2\big(\$alpha^{2}$+(1-\$\alpha$)^3 $e^{{4T}}$+4\$\alpha$(1-\$\alpha$)^2 $e^{{2T}}$\big)}{4(1-\$\alpha$)^3} + O(\hat\$lambda^{3}$),$$ with a parallel expression in (2.8) for $\alpha>1$ that begins with $d(\alpha-1)/(4\alpha)$. The formula implies that once $n>d$, the KL divergence is monotonically decreasing in $n/d$, and that in the $n/d\to\infty$ limit the divergence per dimension scales like $\lambda^2 e^{4T}(1-e^{-2T})^2$, so exact recovery of the underlying distribution requires $\lambda e^{2T}(1-e^{-2T})\ll 1$. For a Gaussian-mixture training set with the convolutional architecture used in practical diffusion models, the same single-step analysis predicts that the final denoising step improves by a factor of $s^2$ relative to a one-step model, explaining empirically observed quality gains from more diffusion steps.

Load-bearing premise

The explicit formulas assume the noise in the denoising regression is independent of the noise added to the training inputs, although the generative model defines them through the same random draw; the paper asserts, without proof, that this mismatch affects only higher-order terms.

Editorial extensions

If this is right

  • The generation gap per dimension becomes a function of the ratio $\alpha=d/n$ alone, so in the proportional regime adding data beyond $n>d$ monotonically shrinks the KL divergence, with a singular divergence at $n=d$.
  • Exact recovery of the sampling distribution requires $\lambda e^{2T}(1-e^{-2T})\ll 1$; finite diffusion time, not just noise scale, sets the achievable fidelity.
  • In a multi-step diffusion model with Gaussian-mixture data, the last denoising step's error scales as $\lambda^2\beta^2(t/s)^2$, giving a factor $s^2$ improvement in the final step as the number of steps grows.
  • The practical criterion $\lambda\beta\ll 1$ emerges for good generation quality, connecting the toy model's solvability to hyperparameter choices in realistic diffusion schedules.

Reading between the lines

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

  • Beyond the paper: the mismatch between the regression model in (C.1) and the generative noise model in (2.2) is testable; recomputing the first-order coefficient with the errors-in-variables regression would show whether the displayed formula is the exact generation gap of the defined model or of a nearby model.
  • Beyond the paper: composing $s$ copies of the linear denoiser should give a closed-form KL gap in $s$ with the same free-product calculus, providing an analytic check of the $s^2$ scaling that is only argued heuristically.
  • Beyond the paper: for non-Gaussian sampling densities with known covariance spectrum, the same deterministic traces would replace $d$ with spectral sums, extending the formula to a wider class of data while keeping the calculation explicit.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 a single-step, linear-denoiser generative model in which noisy samples are produced by an Ornstein-Uhlenbeck-style forward process, a linear regression denoiser is trained on those samples, and the generated distribution is obtained by applying the denoiser to a new noisy sample. The main claimed result is a closed-form, high-dimensional asymptotic expression for the variance part of the KL divergence between the generated and the underlying isotropic Gaussian distributions, together with a phase-transition statement at α = d/n = 1 and an explanation, based on that formula, of why more diffusion steps improve output quality in practical DDPMs. The proof uses deterministic equivalence / random matrix theory in Appendix C, and numerical experiments are reported in Figures 1 and 2.

Significance. If the central derivation were correct, the paper would provide a rare exactly solvable toy model in which the generation gap of a diffusion-like model can be predicted in the n,d → ∞ limit with fixed α = d/n. The claimed explicit formulas (2.7) and (2.8), the monotonicity result for α < 1, and the qualitative explanation of step-count effects in Section 3 would be useful and testable. The paper also connects to a well-established random-matrix toolkit rather than fitting free constants, which is a strength. However, the main theorem is derived for a regression problem that does not match the generative model defined in Section 2, so the headline formulas are not established for the stated model; this makes the significance of the numerical comparison in Figure 1 unclear.

major comments (3)
  1. [Appendix C, Eq. (C.1) and Section 2, Eq. (2.2)] The regression solved in Appendix C is not the regression defined in Section 2. In the generative model, X_k = e^{-T}Y_k + sqrt(ΔT)Z_k with Y_k and Z_k independent, so the pair (X,Y) is jointly Gaussian and the residual Y_k - e^T X_k = -e^T sqrt(ΔT)Z_k is correlated with X_k. The true conditional slope is a_* = e^{-T}σ²/(e^{-2T}σ²+ΔT), with residual covariance σ²ΔT/(e^{-2T}σ²+ΔT)I. Appendix C instead solves Y_k = e^T X_k + Z_k with X_k and Z_k independent, i.e., slope e^T and residual variance ΔT. Footnote 8's moment comparison does not establish equivalence because the least-squares estimator is controlled by the conditional distribution, not by the first two marginal moments. At λhat ≪ 1, a_* - e^T = O(λhat) and the residual covariance is approximately e^{2T}ΔT I, a factor e^{2T} larger than ΔT. Consequently the O(λhat) term of (2.7) is off by a factor e^{2T}: the correct coefficient is d α λhat(1-e^{-2T})/(2(1-α)), not d α λhat e^{-4T}(e^{2T}-1)/(2(1-α)). Thus (2.7), (2.8), Lemma 3, and the comparison in Figure 1 describe a different model than the one defined in Section 2.
  2. [Theorem 1 / Lemma 2] The one-line proof of Lemma 2, 'Derivative of RHS in (2.7) with respect to α is positive', is not a complete verification: it is not stated whether the derivative is taken at fixed d or fixed n, and the derivative is not displayed. Since monotonicity of KLvar in n/d is one of the paper's central qualitative claims, this step needs a full derivation or a clear reference to a verified calculation.
  3. [Appendix C, Eq. (C.10) / Theorem 1] The abstract states that the paper presents 'an explicit formula for the Kullback-Leibler divergence', but Theorem 1 actually computes the expectation of KLvar, which is only a lower bound on KL because KLmean is dropped. The smallness of KLmean is supported only by a numerical regime in Figure 1(a), not by an analytic bound. This is a discrepancy between the stated contribution and the proved statement, and it should be corrected in the presentation.
minor comments (4)
  1. [Section 3, Eq. (3.2)] The application of Lemma 3 to DDPM steps is heuristic: the identification β_t = 1 - e^{-2βt/s} is used to map each step to the linear model, but the argument that 'each step improves' and the implicit neglect of correlations between steps are not justified. The claims would be stronger if the predicted 1/s² scaling were tested quantitatively against the experiments in Figure 2.
  2. [Appendix C, Eqs. (C.24)-(C.28)] The displayed formulas for C_{a,b} and B_a contain branch-dependent expressions with sgn functions, and the final ridgeless results use particular branches without explaining which regime each branch corresponds to. A reader cannot easily verify the final simplification without additional explanation.
  3. [Throughout] There are numerous typographical and notation issues, including inconsistent use of hat symbols (e.g., λhat vs λ), missing parentheses in displayed formulas, and undefined quantities such as 'Thse are in agreement with the plot in figure 1'. These should be cleaned up in any revision.
  4. [Figure 2] The figure legend states that 'we can clearly see that image quality improves as diffusion steps increase', but no quantitative metric or error bars are reported for the image-quality claim. Please provide a precise evaluation protocol.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 1 is a parameter-free high-dimensional asymptotic computation whose formulas are not fitted to the numerical experiments.

full rationale

I walked the derivation chain from the generative definition (2.2), through the least-squares denoiser (2.3), to the Gaussian KL decomposition (2.6) and the deterministic-equivalence calculation in Appendix C. The variance contribution is computed by direct random-matrix asymptotics: equations (C.10), (C.29), and (C.30) express the relevant expectation in terms of d, n, T, lambda, and sigma with no fitted constants, and the cited deterministic-equivalence facts (Appendix B, Atanasov et al. 2024) are external results by other authors rather than a self-citation chain. The small-lambda perturbative assumption and the alpha-to-0 or alpha-to-infinity limits are stated assumptions used to obtain tractable closed forms, not post-hoc fits. The genuine internal weakness is in footnote 8 and equation (C.1): the appendix analyzes the regression Y_k = e^T X_k + Z_k with Z_k independent of X_k, whereas the generative model (2.2) defines X_k = e^{-T}Y_k + sqrt(DeltaT)Z_k, so the residual Y_k - e^T X_k = -e^T sqrt(DeltaT)Z_k is correlated with the design matrix and has variance e^{2T}DeltaT rather than DeltaT. That is a model-consistency defect that changes the numerical coefficients in (2.7)/(2.8), but it is not circularity: the theorem's output is not equal to its input by construction, and no fitted parameter is renamed as a prediction. Therefore the circularity score is 0, and the footnote 8 issue should be weighed in the correctness pass rather than here.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central formula depends on standard random matrix theory (free probability, deterministic equivalence), an isotropic Gaussian data assumption, a small-noise expansion, and a problematic independence assumption in the regression analysis that does not follow from the generative model. Section 3 adds a heuristic noise-schedule mapping to relate the toy result to real diffusion models.

free parameters (2)
  • Noise scale lambda
    Model hyperparameter in eq (2.2) controlling the added noise DeltaT = lambda(1 - e^{-2T}); the theoretical results are a perturbative expansion in the rescaled parameter lambda e^{2T}/sigma^2, and the recovery condition depends on lambda.
  • Diffusion time cutoff T
    Model hyperparameter in eq (2.2) controlling how far the forward process runs; the KL formula and the step-count argument both depend on T.
assumptions (5)
  • standard math Free probability and deterministic equivalence results for invariant random matrices
    Appendix B invokes S-transforms, free products, and the deterministic equivalence principle, following Atanasov et al. (2024); these are unproved background results.
  • domain assumption Sampling distribution is isotropic Gaussian
    Theorem 1 and the explicit formulas cover only rho = N(mu, sigma^2 I_d); the claim does not extend to general data distributions.
  • ad hoc to paper Independence of X_k and Z_k in the regression analysis
    Appendix C eq (C.1) states that X_k and Z_k are taken independent of each other, which contradicts the generative model (2.2) where Z_k enters X_k directly; footnote 8 asserts equivalence to leading order without a derivation.
  • ad hoc to paper Small-noise perturbative regime
    The closed forms (2.7) and (2.8) are second-order expansions in the rescaled noise parameter with the rescaled parameter much smaller than 1; the theorem's lower bound holds generally but the explicit expectation does not.
  • ad hoc to paper Noise schedule mapping from DDPM to linear denoiser
    Section 3 equates a multistep nonlinear DDPM to repeated linear denoisers through the approximate schedule beta_t = 1 - e^{-2 beta t/s}; this mapping is not a proven theorem for the U-Net model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A solvable generative model with a linear, one-step denoiser." pith.science (2026). https://pith.science/paper/6UMFUXUF

@misc{pith2026241117807,
  author       = {Pith},
  title        = {Pith review of: A solvable generative model with a linear, one-step denoiser},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6UMFUXUF}},
  note         = {Machine review of arXiv:2411.17807}
}
read the original abstract

We develop an analytically tractable single-step diffusion model based on a linear denoiser and present an explicit formula for the Kullback-Leibler divergence between the generated and sampling distribution, taken to be isotropic Gaussian, showing the effect of finite diffusion time and noise scale. Our study further reveals that the monotonic fall phase of Kullback-Leibler divergence begins when the training dataset size reaches the dimension of the data points. Finally, for large-scale practical diffusion models, we explain why a higher number of diffusion steps enhances production quality based on the theoretical arguments presented before.

Figures

Figures reproduced from arXiv: 2411.17807 by the authors.

Figure 1
Figure 1. In figure (a), we plot various contributions to KL divergence between the generated [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The plot is based on PyTorch-based implementation of the algorithm in [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Diffusion models under low-noise regime

    cs.CV 2025-06 conditional novelty 6.0 of 10

    Diffusion models trained on disjoint data converge at high noise but diverge near the data manifold, and they fail to denoise very small perturbations accurately.

Reference graph

Works this paper leans on

40 extracted references · 5 canonical work pages · cited by 1 Pith paper

  1. [2]

    Explaining neural scaling laws

    Yasaman Bahri, Ethan Dyer, Jared Kaplan, Jaehoon Lee, and Utkarsh Sharma. Explaining neural scaling laws. arXiv preprint arXiv:2102.06701 ,

    arXiv

  2. [3]

    Blake Bordelon and Cengiz Pehlevan

    URL https://arxiv.org/abs/2308.03686. Blake Bordelon and Cengiz Pehlevan. Self-consistent dynamical field theory of kernel evolution in wide neural networks. Advances in Neural Information Processing Systems , 35:32240–32256,

    arXiv

  3. [5]

    Abdulkadir Canatar, Blake Bordelon, and Cengiz Pehlevan

    URL https://arxiv.org/abs/2408.09000. Abdulkadir Canatar, Blake Bordelon, and Cengiz Pehlevan. Spectral bias and task- model alignment explain generalization in kernel regression and infinitely wide neu- ral networks. Nature Communications , 12(1), May 2021a. ISSN 2041-1723. doi: 10.1038/s41467-021-23103-1. URL http://dx.doi.org/10.1038/s41467-021-23103-1...

    arXiv

  4. [7]

    Muthu Chidambaram, Khashayar Gatmiry, Sitan Chen, Holden Lee, and Jianfeng Lu

    URL https://arxiv.org/abs/2301.10972. Muthu Chidambaram, Khashayar Gatmiry, Sitan Chen, Holden Lee, and Jianfeng Lu. What does guidance do? a fine-grained analysis in a simple setting,

    arXiv

  5. [8]

    St´ ephane d’Ascoli, Levent Sagun, and Giulio Biroli

    URL https: //arxiv.org/abs/2409.13074. St´ ephane d’Ascoli, Levent Sagun, and Giulio Biroli. Triple descent and the two kinds of overfitting: Where & why do they appear? Advances in Neural Information Processing Systems, 33:3058–3069,

    arXiv

  6. [9]

    Prafulla Dhariwal and Alex Nichol

    URL https://arxiv.org/abs/2307.03223. Prafulla Dhariwal and Alex Nichol. Diffusion models beat gans on image synthesis,

    arXiv

  7. [10]

    Oussama Dhifallah and Yue M Lu

    URL https://arxiv.org/abs/2105.05233. Oussama Dhifallah and Yue M Lu. A precise performance analysis of learning with random features. arXiv preprint arXiv:2008.11904 ,

    arXiv 2008

  8. [15]

    Universality laws for high-dimensional learning with random features

    Hong Hu and Yue M Lu. Universality laws for high-dimensional learning with random features. IEEE Transactions on Information Theory , 69(3):1932–1964,

    work page 1932

Show all 40 references
  1. [16]

    Arthur Jacot, Franck Gabriel, and Clement Hongler

    URL https://arxiv.org/abs/2212.11972. Arthur Jacot, Franck Gabriel, and Clement Hongler. Neural tangent kernel: Convergence and generalization in neural networks. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors, Advances in Neural ...

    arXiv

  2. [17]

    Arthur Jacot, Franck Gabriel, and Cl´ ement Hongler

    URL https://proceedings.neurips.cc/ paper_files/paper/2018/file/5a4be1fa34e62bb8a6ec6b91d2462f5a-Paper.pdf. Arthur Jacot, Franck Gabriel, and Cl´ ement Hongler. Neural tangent kernel: Convergence and generalization in neural networks,

    2018

  3. [18]

    Zahra Kadkhodaie and Eero Peter Simoncelli

    URL https://arxiv.org/abs/1806.07572. Zahra Kadkhodaie and Eero Peter Simoncelli. Solving linear inverse problems using the prior implicit in a denoiser. In NeurIPS 2020 Workshop on Deep Learning and Inverse Problems,

    2020 arXiv

  4. [19]

    Zahra Kadkhodaie, Florentin Guth, Eero P Simoncelli, and St´ ephane Mallat

    URL https://openreview.net/forum?id=RLN7K4U3UST. Zahra Kadkhodaie, Florentin Guth, Eero P Simoncelli, and St´ ephane Mallat. Generalization in diffusion models arises from geometry-adaptive harmonic representations. arXiv preprint arXiv:2310.02557,

    arXiv

  5. [20]

    An analytic theory of creativity in convolutional diffusion models

    Mason Kamb and Surya Ganguli. An analytic theory of creativity in convolutional diffusion models. arXiv preprint arXiv:2412.20292 ,

    arXiv

  6. [21]

    Anders Krogh and John A Hertz

    URL https://arxiv.org/abs/2206.00364. Anders Krogh and John A Hertz. Generalization in a linear perceptron in the presence of noise. Journal of Physics A: Mathematical and General , 25(5):1135,

    arXiv

  7. [22]

    Jaehoon Lee, Lechao Xiao, Samuel Schoenholz, Yasaman Bahri, Roman Novak, Jascha Sohl-Dickstein, and Jeffrey Pennington

    URL https://arxiv.org/abs/2206.06227. Jaehoon Lee, Lechao Xiao, Samuel Schoenholz, Yasaman Bahri, Roman Novak, Jascha Sohl-Dickstein, and Jeffrey Pennington. Wide neural networks of any depth evolve as linear models under gradient descent. In H. Wallach, H. Larochelle, A. Beyg...

    arXiv

  8. [24]

    A solvable model of neural scaling laws

    Alexander Maloney, Daniel A Roberts, and James Sully. A solvable model of neural scaling laws. arXiv preprint arXiv:2210.16859 ,

    arXiv

  9. [25]

    doi: 10.1073/pnas.1806579115

    ISSN 1091-6490. doi: 10.1073/pnas.1806579115. URL http://dx.doi.org/ 10.1073/pnas.1806579115. Song Mei, Theodor Misiakiewicz, and Andrea Montanari. Generalization error of random feature and kernel methods: hypercontractivity and kernel matrix concentration. Applied and Comput...

    doi

  10. [26]

    More data can hurt for linear regression: Sample-wise double descent

    Preetum Nakkiran. More data can hurt for linear regression: Sample-wise double descent. arXiv preprint arXiv:1912.07242 ,

    1912 arXiv

  11. [27]

    Maya Okawa, Ekdeep Singh Lubana, Robert P

    URL https://arxiv.org/abs/2102.09672. Maya Okawa, Ekdeep Singh Lubana, Robert P. Dick, and Hidenori Tanaka. Compositional abilities emerge multiplicatively: Exploring diffusion models on a synthetic task. arXiv preprint arXiv:2310.09336,

    arXiv

  12. [28]

    Daniel A

    URL https://arxiv.org/abs/ 2204.06125. Daniel A. Roberts, Sho Yaida, and Boris Hanin. The Principles of Deep Learning Theory: An Effective Theory Approach to Understanding Neural Networks . Cambridge University Press, May

    arXiv

  13. [29]

    doi: 10.1017/9781009023405

    ISBN 9781316519332. doi: 10.1017/9781009023405. URL http: //dx.doi.org/10.1017/9781009023405. Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨ orn Ommer. High-resolution image synthesis with latent diffusion models,

    doi

  14. [30]

    org/abs/2112.10752

    URL https://arxiv. org/abs/2112.10752. 21 Halder Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation,

    arXiv

  15. [32]

    Kulin Shah, Sitan Chen, and Adam Klivans

    URL https: //arxiv.org/abs/2205.11487. Kulin Shah, Sitan Chen, and Adam Klivans. Learning mixtures of gaussians using the ddpm objective,

    arXiv

  16. [33]

    James B Simon, Madeline Dickens, Dhruva Karkada, and Michael Deweese

    URL https://arxiv.org/abs/2307.01178. James B Simon, Madeline Dickens, Dhruva Karkada, and Michael Deweese. The eigenlearning framework: A conservation law perspective on kernel ridge regression and wide neural networks. Transactions on Machine Learning Research,

    arXiv

  17. [34]

    org/abs/1503.03585

    URL https://arxiv. org/abs/1503.03585. Peter Sollich. Learning curves for Gaussian processes. Advances in neural information processing systems, 11,

    arXiv

  18. [35]

    Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole

    Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations, 2021b. Stefano Spigler, Mario Geiger, and Matthieu Wyart. Asymptotic learning curves of ker- nel metho...

    2020

  19. [37]

    Greg Yang and Edward J

    URL https: //arxiv.org/abs/2403.01639. Greg Yang and Edward J. Hu. Tensor programs iv: Feature learning in infinite-width neural networks. In Marina Meila and Tong Zhang, editors, Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings o...

    arXiv

  20. [38]

    TaeHo Yoon, Joo Young Choi, Sehyun Kwon, and Ernest K

    URL https://proceedings.mlr.press/ v139/yang21c.html. TaeHo Yoon, Joo Young Choi, Sehyun Kwon, and Ernest K. Ryu. Diffusion probabilistic models generalize when they fail to memorize. In ICML 2023 Workshop on Structured Probabilistic Inference & Generative Modeling ,

    2023

  21. [39]

    The emergence of reproducibility and generalizability in diffusion models

    Huijie Zhang, Jinfan Zhou, Yifu Lu, Minzhe Guo, Peng Wang, Liyue Shen, and Qing Qu. The emergence of reproducibility and generalizability in diffusion models. arXiv preprint arXiv:2310.05264,

    arXiv

  22. [40]

    org/abs/2110.07579

    URL https://arxiv. org/abs/2110.07579. 23

    arXiv

  23. [2011]

    22 A solvable generative model with a linear, one-step denoiser Yuchen Wu, Minshuo Chen, Zihao Li, Mengdi Wang, and Yuting Wei

    doi: 10.1162/NECO a 00142. 22 A solvable generative model with a linear, one-step denoiser Yuchen Wu, Minshuo Chen, Zihao Li, Mengdi Wang, and Yuting Wei. Theoretical insights for diffusion guidance: A case study for gaussian mixture models,

    doi

  24. [2015]

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S

    URL https://arxiv.org/abs/1505.04597. Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S. Sara Mahdavi, Rapha Gontijo Lopes, Tim Salimans, Jonathan Ho, David J Fleet, and Mohammad Norouzi. Phot...

    arXiv

  25. [2016]

    URL https://doi.org/10.3150/14-BEJ609

    doi: 10.3150/14-BEJ609. URL https://doi.org/10.3150/14-BEJ609. Edgar Dobriban and Stefan Wager. High-dimensional asymptotics of prediction: Ridge regression and classification. The Annals of Statistics , 46(1):247 – 279,

    doi

  26. [2018]

    URL https://doi.org/10.1214/17-AOS1549

    doi: 10.1214/17-AOS1549. URL https://doi.org/10.1214/17-AOS1549. Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An ...

    doi

  27. [2019]

    Marvin Li and Sitan Chen

    URL https://proceedings.neurips.cc/ paper_files/paper/2019/file/0d1a9651497a38d8b1c3871c84528bd4-Paper.pdf. Marvin Li and Sitan Chen. Critical windows: non-asymptotic theory for feature emergence in diffusion models. In International Conference on Machine Learning , pages 2747...

    2019

  28. [2020]

    mlr.press/v119/d-ascoli20a.html

    URL http://proceedings. mlr.press/v119/d-ascoli20a.html. ISSN: 2640-3498. Alessandro Favero, Antonio Sclocchi, Francesco Cagnetta, Pascal Frossard, and Matthieu Wyart. How compositional generalization and creativity improve as diffusion models are trained. arXiv preprint arXiv...

    arXiv

  29. [2021]

    Arwen Bradley and Preetum Nakkiran

    URL https://arxiv.org/ abs/2002.02561. Arwen Bradley and Preetum Nakkiran. Classifier-free guidance is a predictor-corrector,

    2002 arXiv

  30. [2022]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel

    URL https: //arxiv.org/abs/2207.12598. Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems , 33:6840–6851,

    arXiv

  31. [2023]

    Improved analysis of score-based generative modeling: User-friendly bounds under minimal smoothness assumptions, 2023a

    Hongrui Chen, Holden Lee, and Jianfeng Lu. Improved analysis of score-based generative modeling: User-friendly bounds under minimal smoothness assumptions, 2023a. URL https://arxiv.org/abs/2211.01916. Sitan Chen, Sinho Chewi, Jerry Li, Yuanzhi Li, Adil Salim, and Anru R. Zhang...

    arXiv

  32. [2024]

    Francis Bach

    URL https://arxiv.org/abs/2405.00592. Francis Bach. High-dimensional analysis of double descent for linear regression with random projections. SIAM Journal on Mathematics of Data Science , 6(1):26–50,

    arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.