Pith. sign in

REVIEW 3 major objections 5 minor 44 references

CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation

T0 review · 3 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read DDIM diffusion sampling responds almost linearly to initial-noise perturbations, and the paper builds a sampling controller on that fact.

desk verdict A genuinely useful empirical linearity result, but the paper's central mean-control claim rests on an unjustified expectation step that treats a random Jacobian as constant. read the letter →

arxiv 2502.04670 v1 pith:2TBH3TKO submitted 2025-02-07 cs.LG cs.AI

classification cs.LGcs.AI
keywords diffusionmodelsDDIMsamplinginitialnoiseperturbationlinearresponsecontrollablegenerationconstrainedsphericalinterpolationsamplediversity
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

Diffusion sampling, in the deterministic DDIM form, has a simple response to small changes in the random noise it starts from: the final image moves in a nearly linear way. The paper proves an expansion $x_0(x_T + \lambda \Delta x, T) = x_0(x_T) + \lambda \gamma_0(x_T) \Delta x + o(\lambda)$ using only smoothness of the log-density, with an analogous statement for the probability-flow ODE, and measures near-perfect linearity ($R^2 \ge 0.94$) on four image benchmarks. On top of this, the proposed CCS algorithm interpolates the original noise with a fresh Gaussian vector on the sphere, and binary-searches the interpolation angle so the average distance from a target image lands on a user-specified diversity level. If the linearity result holds, this gives a training-free way to enforce distribution-level constraints on diffusion samples, such as a target mean and a chosen spread, without per-sample projections that degrade quality.

What carries the argument

The object carrying the argument is the Jacobian $\gamma_0(x_T)$ of the map from initial noise to final sample. In the discrete setting it is built by composing one-step DDIM maps $L_t(x) = \eta_t x + \lambda_t \nabla \log p_t(x)$ and differentiating: $\gamma_T(x) = \eta_T + \lambda_T H_T(x)$, with $\gamma_{t-1}$ obtained by the chain rule through the recursion; smoothness of the log-density, namely existence of the Hessian, is the sufficient condition. The same Jacobian emerges in the continuous ODE from differentiability of the flow with respect to its initialization, proved with Picard-Lindelof and Gronwall arguments. The CCS algorithm then uses the spherical interpolation formula $x'_T = \frac{\sin(C_0)}{\sin(\theta)} \epsilon + \frac{\sin(\theta - C_0)}{\sin(\theta)} x_T$ with $\theta = \angle(\epsilon, x_T)$, which preserves the Gaussian-sphere norm, and a binary-search controller that tunes $C_0$ to hit a target MSE; the linear response is what makes the binary search converge quickly.

What would settle it

Run the CCS algorithm around one fixed target image with many random draws at several $C_0$ values and compute the empirical sample mean of the generated images: if its distance from the target grows with $C_0$ beyond the controller tolerance (equivalently, if $E[\gamma_0(a x_T + b\Delta)\Delta]$ is not small), the centering claim fails.

Watch

Extended reading notes

Core claim

The central discovery claimed is that the input-output map of DDIM/ODE diffusion sampling is locally linear in the initial noise: Proposition 1 gives $x_0(x_T + \lambda \Delta x, T) = x_0(x_T) + \lambda \gamma_0(x_T) \Delta x + o(\lambda)$, and Proposition 2 gives the continuous-time analogue. The authors then claim this linearity can be exploited: by writing $x'_T = \frac{\sin(C_0)}{\sin(\theta)} \epsilon + \frac{\sin(\theta - C_0)}{\sin(\theta)} x_T$, the generated sample keeps high quality because the perturbed noise stays on the Gaussian hypersphere, while its mean stays near the target $x_0$; the binary-search controller picks $C_0$ so that $E[\|x'_0 - x_0\|_2]$ equals a target MSE. Empirically, the normalized average residual norm tracks $\sin(C_0)$ with $R^2$ values of 0.995 on FFHQ, 0.988 on CIFAR-10, 0.959 on CelebA-HQ, and 0.947 on fMoW, and CCS achieves the highest PSNR among the compared controller-based methods on all three image benchmarks.

Load-bearing premise

When the paper averages over random perturbations, it treats the sampler's local linear response as a fixed multiplier independent of the perturbation; if that multiplier correlates with the perturbation, the sample mean will drift off the target, and the paper neither proves nor tests that correlation is zero.

Editorial extensions

If this is right

  • A user can specify a target image and a desired diversity level (rMSE), and the controller finds a perturbation angle $C_0$ such that the generated batch has that average distance while its sample mean stays close to the target.
  • The spherical-interpolation mechanism avoids the quality collapse that plain additive noise causes: Proposition 5 shows additive perturbations push the input off the Gaussian hypersphere, while Proposition 4 guarantees a standard Gaussian noise vector stays on it.
  • Because the linearity also holds, more weakly, for latent diffusion models through partial inversion, the method applies to text-guided generators and to controlled image editing between a source prompt and a target prompt.
  • The reported $R^2$ values, 0.995 on FFHQ, 0.988 on CIFAR-10, 0.959 on CelebA-HQ, and 0.947 on fMoW, quantify how predictable the distance control is on these benchmarks.

Reading between the lines

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

  • Beyond the paper: if the noise-to-image map is linear in this regime, then distribution-level constraints beyond a target mean, such as a target covariance or marginal moments, could be imposed by choosing the distribution of the perturbation $\Delta$ and pushing it through the estimated Jacobian, without per-sample projections.
  • Beyond the paper: the same Jacobian $\gamma_0$ could be estimated once per target image and reused to solve multi-constraint control problems with a single linear solve, replacing the binary-search controller.
  • Beyond the paper: the lower $R^2$ on the latent diffusion benchmarks suggests the nonlinear autoencoder and inexact classifier-free-guidance inversion are the main sources of nonlinearity; testing the same spherical perturbation in a better-inverted latent space would isolate whether the linearity is a property of the ODE sampler or of the full pipeline.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper studies how perturbing the initial noise vector of DDIM/ODE sampling affects the generated image. It claims, both theoretically (Propositions 1-3) and empirically (Table 1), that the output changes approximately linearly with the perturbation scale. Based on this, it proposes CCS (and P-CCS for latent diffusion), which uses spherical interpolation between the inverted noise of a target mean and a random Gaussian vector, with a binary-search controller (Algorithm 2) to match a desired output diversity (rMSE). Experiments on FFHQ, CIFAR-10, CelebA-HQ, and fMoW report that CCS achieves higher PSNR between sample mean and target mean than adapted baselines while preserving quality and diversity. The paper also demonstrates an image-editing application.

Significance. If the linearity and centering claims are correct, CCS is a simple, training-free tool for distribution-level control of diffusion sampling, relevant to privacy, constrained generation, and editing. The reported linearity phenomenon (R^2 between 0.947 and 0.995 across four datasets) is a useful empirical observation, and the theoretical first-order expansion (Proposition 1) is a reasonable pointwise statement. However, the paper's main contribution over prior per-sample control is the claim that E[x'_0] ≈ x0, and that claim rests on a derivation gap (Section 4.1) and aggregate statistics without per-target residuals or confidence intervals. The method itself is novel and the empirical results are promising, but the central guarantee needs a corrected derivation or direct empirical validation before publication.

major comments (3)
  1. [Section 4.1] The derivation of E[x'_0] ≈ x0 is invalid as written. The paper defines A = γ0(a xT + bΔ) and then writes E[x̂'_0] = x0 + b A E[Δ], treating A as a constant when taking the expectation. But A is a function of the random perturbation Δ, so the second term should be b E[γ0(a xT + bΔ)Δ]. The paper neither proves that this expectation vanishes nor reports an experiment that isolates this term. This step is load-bearing for the 'close to target mean' design goal, so the central claim is not established by the provided theory.
  2. [Section 4.1 and Algorithm 1] Even if A were treated as constant, the perturbation is x'_T − x_T = (a−1)x_T + bΔ, with a = sin(θ−C0)/sin(θ) and b = sin(C0)/sin(θ). For C0 > 0, a < 1, so the first-order output shift includes (a−1) γ0(x_T) x_T, which is not zero. Proposition 1 is a pointwise expansion in a scalar λ and does not justify dropping this radial term or moving the Jacobian outside the expectation. Consequently, the conclusion E[x'_0] ≈ x0 in Section 4.1 is not a consequence of the stated propositions.
  3. [Section 5.2 and Proposition 6] Proposition 6 controls only the input distance ||x'_T − x_T||, not the output mean. The bridge from input-distance control to output-rMSE control relies on the empirically observed linearity, but the reported PSNR values in Tables 2-4 are averaged over all target means without per-target residuals, confidence intervals, or ablations over C0. Since the theoretical centering argument is incomplete, the empirical evidence should directly test the cancellation E[γ0(a xT + bΔ)Δ] ≈ 0 and the contribution of the (a−1) term; currently it does not.
minor comments (5)
  1. [Algorithm 2] The binary-search update rules in lines 8 and 10 appear to be mistyped: they read '(Chigh, C0) ← (C0, C0+Clow/2)' and '(Clow, C0) ← (C0, C0+Chigh/2)', which are ambiguous and would not set C0 to the midpoint of the new interval. Please rewrite the updates clearly.
  2. [Proposition 4] The probability statement in Proposition 4 has a misplaced bracket: it should read P(||X||_2^2 ∈ (1±δ)d) ≥ 1 − 2 exp(...), not P(||X||_2^2 ∈ (1±δ)d]).
  3. [Section 5.1] The R^2 values in Table 1 are computed after per-target normalization y' = (y−b)/a using a fitted slope and intercept. This normalization removes the per-target slope information and can inflate the apparent linearity; please also report R^2 on the unnormalized residuals or the distribution of per-target slopes.
  4. [Algorithm 3] In Step 2 of Algorithm 3, the sampling notation 'ϵ ∼ N(0, 1 − αt)' should specify the variance as (1 − α_{t0})I with matching dimensionality; as written it is ambiguous.
  5. [General] No code or reproducible implementation details are provided, which makes it difficult to verify the exact inversion procedures, especially the partial inversion for latent diffusion (P-CCS) where t0 = 45 is a tuned hyperparameter on one validation image.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; the core linearity is proved via ODE stability and the control loop is feedback, with only a non-circular derivation gap in Section 4.1.

full rationale

Walking the derivation chain: Proposition 1 is proved in Appendix A.1 by composing the DDIM recursion and Taylor-expanding the score; Propositions 2 and 3 follow from ODE-flow differentiability and Grönwall's inequality. These results do not use CCS or fit any parameter to the data. Table 1's R2 is a goodness-of-fit measure: per-target affine normalization does not change R2 under the fitted line, so it is not a construction artifact, and the fitted slope/intercept are not reused as predictions. Algorithm 2 is a feedback controller (binary search on measured output MSE), not a fitted prediction. The spherical-interpolation distance bound (Proposition 6) is a geometric calculation. The self-citations (Song et al. 2023 as a baseline; Wang et al. 2024 as related work) are not load-bearing; the central derivation and benchmarks are self-contained. Section 4.1 does contain a genuine gap: with A = γ0(axT + bΔ), the paper writes E[x̂'0] = x0 + bA E[Δ], treating A as non-random, and it drops the radial term (a−1)xT. But E[AΔ] ≠ A E[Δ] in general, so the mean-centering claim is not established. This is a correctness/derivation gap, not a circular reduction: no equation is made equal to its own input by definition, and no fitted parameter is renamed a prediction.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central results rest on smoothness of the score, high-dimensional Gaussian concentration, and an unproven independence or correlation condition for the Jacobian inside an expectation. No new physical entities are introduced. One hyperparameter, t0, is tuned on validation data for the latent diffusion variant.

free parameters (1)
  • partial inversion timestep t0 for latent diffusion P-CCS = 45 (with T = 50)
    Chosen in Appendix B.1 by tuning on one held-out validation target image; affects the quality-diversity tradeoff of P-CCS and is not derived from theory.
assumptions (4)
  • domain assumption The score function log p_t is C^2, or h(t, y) is C^1 and Lipschitz in y
    Used in Propositions 1, 2 and 3 to justify the Taylor expansion and ODE stability; standard for idealized diffusion models but not verified for real neural score networks.
  • domain assumption The DDIM-inverted noise x_T and fresh Gaussian noise have norm (1 + o(1)) sqrt(d) and are nearly orthogonal in high dimension
    Used in Proposition 4 and 6 and Algorithm 1; Gaussian concentration is standard, but x_T coming from DDIM inversion is assumed to behave like random Gaussian noise.
  • ad hoc to paper The Jacobian A = gamma0(a xT + b Delta) is treated as constant when computing E[x'_0] = x0 + b A E[Delta]
    Section 4.1 requires neglecting or canceling E[A Delta]; this is not proved and is the main unverified step linking input noise control to output mean control.
  • domain assumption The neural network score is accurate enough for the idealized DDIM process and for exact inversion in the unconditional pixel-diffusion setting
    The theory replaces the learned noise predictor with the true score; real pretrained models introduce approximation error, which the paper acknowledges for latent and classifier-free-guidance settings by switching to partial inversion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation." pith.science (2026). https://pith.science/paper/2TBH3TKO

@misc{pith2026250204670,
  author       = {Pith},
  title        = {Pith review of: CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2TBH3TKO}},
  note         = {Machine review of arXiv:2502.04670}
}
read the original abstract

Diffusion models have emerged as powerful tools for generative tasks, producing high-quality outputs across diverse domains. However, how the generated data responds to the initial noise perturbation in diffusion models remains under-explored, which hinders understanding the controllability of the sampling process. In this work, we first observe an interesting phenomenon: the relationship between the change of generation outputs and the scale of initial noise perturbation is highly linear through the diffusion ODE sampling. Then we provide both theoretical and empirical study to justify this linearity property of this input-output (noise-generation data) relationship. Inspired by these new insights, we propose a novel Controllable and Constrained Sampling method (CCS) together with a new controller algorithm for diffusion models to sample with desired statistical properties while preserving good sample quality. We perform extensive experiments to compare our proposed sampling approach with other methods on both sampling controllability and sampled data quality. Results show that our CCS method achieves more precisely controlled sampling while maintaining superior sample quality and diversity.

Figures

Figures reproduced from arXiv: 2502.04670 by the authors.

Figure 1
Figure 1. CCS Sampling with a target mean in the FFHQ validation dataset with C0 = 0.4 Based on the spherical interpolation to perturb the initial noise vector, we propose a novel controllable and con￾strained sampling method (CCS) together with a new con￾troller algorithm for diffusion models to sample with desired statistical properties while preserving high quality and ad￾justable diversity. Furthermore, we conduct extensi… view at source ↗
Figure 2
Figure 2. Qualitative demonstration of linearity when increasing scale of perturbation. For each target mean, we sample a perturbation noise and gradually increase C0 (0.1 at a time) to increase the magnitude of the perturbation. 3.2. DDIM Discretization We simplify notations and write the idealized DDIM (3) as xt−1 := ηtxt + λt∇xt log pt(xt), where ηt = q α −1 t−1αt and λt = q α −1 t−1αt(1 − αt) − p (1 − αt−1)(1 − αt). Fix T… view at source ↗
Figure 3
Figure 3. Quantitative demonstration of linearity when increasing scale of perturbation. With increased sin(C0), the magnitude of perturbation increases, and the average L 2 distance between samples and the target image increases linearly. Left is the linearity on FFHQ dataset using pixel diffusion; Right is the linearity on Celeba-HQ dataset using Stable Diffusion 1.5 [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: We sample 120 images with a fixed target mean using different methods and analyze their sample mean (average pixel intensity). Our observations show that the sample mean of our method closely matches that of the original image [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Example Samples of P-CCS around an Edited Mean. 7 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: CCS-CT Sampled Images with Stable Diffusion 1.5. (a) Samples with an in-the-wild target mean (e) and a target rMSE 0.09; (b) Samples with a target mean (d) from Celeba-HQ dataset and a target rMSE 0.07; (d): sample mean from (a); (f): sample mean from (b) [PITH_FULL_I…
Figure 7
Figure 7. Figure 7: ,8 demonstrates an example of image editing controlled sampling with Alg. 4 [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Image Editing Samples with Stable Diffusion 1.5, the source prompt is given by ‘a high-quality portrait of an old man’, and the target prompt is given by ‘a high-quality portrait of a young man’, the target MSE level is given by 0.09 [PITH_FULL_IMAGE:figures/full_fig_…
Figure 9
Figure 9. Figure 9: We first sample 50 images around each target mean in the FFHQ dataset. We then obtain the DDIM inverse of each target mean, and then add spherical perturbation to it. When the scale of perturbations sin(C0) increases, the average of norms of the residuals between each …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 25 canonical work pages

  1. [1]

    Exploring low-dimensional subspace in diffusion models for controllable image editing

    Chen, S., Zhang, H., Guo, M., Lu, Y., Wang, P., and Qu, Q. Exploring low-dimensional subspace in diffusion models for controllable image editing. In The Thirty-eighth Annual Conference on Neural Information Processing Systems

  2. [2]

    Learning general gaussian mixtures with efficient score matching

    Chen, S., Kontonis, V., and Shah, K. Learning general gaussian mixtures with efficient score matching. arXiv preprint arXiv:2404.18893, 2024 a

  3. [3]

    Exploring low-dimensional subspace in diffusion models for controllable image editing

    Chen, S., Zhang, H., Guo, M., Lu, Y., Wang, P., and Qu, Q. Exploring low-dimensional subspace in diffusion models for controllable image editing. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024 b . URL https://openreview.net/forum?id=50aOEfb2km

  4. [4]

    Ilvr: Conditioning method for denoising diffusion probabilistic models

    Choi, J., Kim, S., Jeong, Y., Gwon, Y., and Yoon, S. Ilvr: Conditioning method for denoising diffusion probabilistic models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 14367--14376, 2021

  5. [5]

    Functional map of the world

    Christie, G., Fendley, N., Wilson, J., and Mukherjee, R. Functional map of the world. In CVPR, 2018

  6. [6]

    T., Klasky, M

    Chung, H., Kim, J., Mccann, M. T., Klasky, M. L., and Ye, J. C. Diffusion posterior sampling for general noisy inverse problems. In The Eleventh International Conference on Learning Representations

  7. [7]

    Chung, H., Sim, B., Ryu, D., and Ye, J. C. Improving diffusion models for inverse problems using manifold constraints. Advances in Neural Information Processing Systems, 35: 0 25683--25696, 2022 a

  8. [8]

    Chung, H., Sim, B., and Ye, J. C. Come-closer-diffuse-faster: Accelerating conditional diffusion models for inverse problems through stochastic contraction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 12413--12422, 2022 b

Show all 44 references
  1. [9]

    C., Milanfar, P., and Delbracio, M

    Chung, H., Ye, J. C., Milanfar, P., and Delbracio, M. Prompt-tuning latent diffusion models for inverse problems. arXiv preprint arXiv:2310.01110, 2023

  2. [10]

    Style injection in diffusion: A training-free approach for adapting large-scale diffusion models for style transfer

    Chung, J., Hyun, S., and Heo, J.-P. Style injection in diffusion: A training-free approach for adapting large-scale diffusion models for style transfer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 8795--8805, 2024

  3. [11]

    Convergence of denoising diffusion models under the manifold hypothesis

    De Bortoli, V. Convergence of denoising diffusion models under the manifold hypothesis. Transactions on Machine Learning Research, 2022

  4. [12]

    and Nichol, A

    Dhariwal, P. and Nichol, A. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34: 0 8780--8794, 2021 a

  5. [13]

    and Nichol, A

    Dhariwal, P. and Nichol, A. Diffusion models beat GAN s on image synthesis. Advances in neural information processing systems, 34: 0 8780--8794, 2021 b

  6. [14]

    Differential privacy

    Dwork, C. Differential privacy. In International colloquium on automata, languages, and programming, pp.\ 1--12. Springer, 2006

  7. [15]

    Diffusion self-guidance for controllable image generation

    Epstein, D., Jabri, A., Poole, B., Efros, A., and Holynski, A. Diffusion self-guidance for controllable image generation. Advances in Neural Information Processing Systems, 36: 0 16222--16239, 2023

  8. [16]

    Learning mixtures of gaussians using diffusion models

    Gatmiry, K., Kelner, J., and Lee, H. Learning mixtures of gaussians using diffusion models. arXiv preprint arXiv:2404.18869, 2024

  9. [17]

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

    Guo, X., Liu, J., Cui, M., Li, J., Yang, H., and Huang, D. Initno: Boosting text-to-image diffusion models via initial noise optimization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 9380--9389, 2024

  10. [18]

    Ordinary differential equations

    Hartman, P. Ordinary differential equations. SIAM, 2002

  11. [19]

    Z., Salakhutdinov, R., et al

    He, Y., Murata, N., Lai, C.-H., Takida, Y., Uesaka, T., Kim, D., Liao, W.-H., Mitsufuji, Y., Kolter, J. Z., Salakhutdinov, R., et al. Manifold preserving guided diffusion. In The Twelfth International Conference on Learning Representations

  12. [20]

    Y., Bae, H., and Chun, S

    Hong, S., Lee, K., Jeon, S. Y., Bae, H., and Chun, S. Y. On exact inversion of dpm-solvers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 7069--7078, 2024

  13. [21]

    A style-based generator architecture for generative adversarial networks

    Karras, T. A style-based generator architecture for generative adversarial networks. arXiv preprint arXiv:1812.04948, 2019

  14. [22]

    Musiq: Multi-scale image quality transformer

    Ke, J., Wang, Q., Wang, Y., Milanfar, P., and Yang, F. Musiq: Multi-scale image quality transformer. In Proceedings of the IEEE/CVF international conference on computer vision, pp.\ 5148--5157, 2021

  15. [23]

    Diffwave: A versatile diffusion model for audio synthesis

    Kong, Z., Ping, W., Huang, J., Zhao, K., and Catanzaro, B. Diffwave: A versatile diffusion model for audio synthesis. In International Conference on Learning Representations

  16. [24]

    Learning multiple layers of features from tiny images

    Krizhevsky, A., Hinton, G., et al. Learning multiple layers of features from tiny images. 2009

  17. [25]

    and Ye, J

    Kwon, T. and Ye, J. C. Solving video inverse problems using image diffusion models. arXiv preprint arXiv:2409.02574, 2024

  18. [26]

    Drag your noise: Interactive point-based editing via diffusion semantic propagation

    Liu, H., Xu, C., Yang, Y., Zeng, L., and He, S. Drag your noise: Interactive point-based editing via diffusion semantic propagation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 6743--6752, 2024

  19. [27]

    S., Agarwal, S., Rout, L., Shakkottai, S., and Chinchali, S

    Narasimhan, S. S., Agarwal, S., Rout, L., Shakkottai, S., and Chinchali, S. P. Constrained posterior sampling: Time series generation with hard constraints. arXiv preprint arXiv:2410.12652, 2024

  20. [28]

    Nichol, A. Q. and Dhariwal, P. Improved denoising diffusion probabilistic models. In International conference on machine learning, pp.\ 8162--8171. PMLR, 2021

  21. [29]

    High-resolution image synthesis with latent diffusion models

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., and Ommer, B. 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

  22. [30]

    Improved techniques for training gans

    Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., and Chen, X. Improved techniques for training gans. Advances in neural information processing systems, 29, 2016

  23. [31]

    Animating rotation with quaternion curves

    Shoemake, K. Animating rotation with quaternion curves. In Proceedings of the 12th annual conference on Computer graphics and interactive techniques, pp.\ 245--254, 1985

  24. [32]

    M., Zhang, Z., Hu, X., Qu, Q., and Shen, L

    Song, B., Kwon, S. M., Zhang, Z., Hu, X., Qu, Q., and Shen, L. Solving inverse problems with latent diffusion models via hard data consistency. In The Twelfth International Conference on Learning Representations

  25. [33]

    Denoising diffusion implicit models

    Song, J., Meng, C., and Ermon, S. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020

  26. [34]

    Denoising diffusion implicit models

    Song, J., Meng, C., and Ermon, S. Denoising diffusion implicit models. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=St1giarCHLP

  27. [35]

    Loss-guided diffusion models for plug-and-play controllable generation

    Song, J., Zhang, Q., Yin, H., Mardani, M., Liu, M.-Y., Kautz, J., Chen, Y., and Vahdat, A. Loss-guided diffusion models for plug-and-play controllable generation. In International Conference on Machine Learning, pp.\ 32483--32498. PMLR, 2023

  28. [36]

    and Ermon, S

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

  29. [37]

    C., and Loy, C

    Wang, J., Chan, K. C., and Loy, C. C. Exploring clip for assessing the look and feel of images. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp.\ 2555--2563, 2023

  30. [38]

    Diffusion models learn low-dimensional distributions via subspace clustering

    Wang, P., Zhang, H., Zhang, Z., Chen, S., Ma, Y., and Qu, Q. Diffusion models learn low-dimensional distributions via subspace clustering. arXiv preprint arXiv:2409.02426, 2024

  31. [39]

    Wu, C. H. and De la Torre, F. A latent space of stochastic diffusion models for zero-shot image editing and guidance. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 7378--7387, 2023

  32. [40]

    Tedigan: Text-guided diverse face image generation and manipulation

    Xia, W., Yang, Y., Xue, J.-H., and Wu, B. Tedigan: Text-guided diverse face image generation and manipulation. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021

  33. [41]

    On the robustness of diffusion inversion in image manipulation

    Zhang, J., Das, K., and Kumar, S. On the robustness of diffusion inversion in image manipulation. In ICLR 2023 Workshop on Trustworthy and Reliable Large-Scale Machine Learning Models, 2023 a

  34. [42]

    Adding conditional control to text-to-image diffusion models

    Zhang, L., Rao, A., and Agrawala, M. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 3836--3847, 2023 b

  35. [43]

    Noisediffusion: Correcting noise for image interpolation with diffusion models beyond spherical linear interpolation

    Zheng, P., Zhang, Y., Fang, Z., Liu, T., Lian, D., and Han, B. Noisediffusion: Correcting noise for image interpolation with diffusion models beyond spherical linear interpolation. In The Twelfth International Conference on Learning Representations, 2024

  36. [44]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.