Pith. sign in

REVIEW 4 major objections 3 minor 62 references

This paper claims that a measurement-residual term can replace the variance-compensation term in a consistency-model sampler, yielding a test-time-only guidance mechanism that improves inverse-problem reconstructions in two steps without re

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 12:41 UTC pith:M4W4NWJU

load-bearing objection A useful, empirically solid sampler variant, but the paper never isolates whether the measurement-residual term—not just extra stochasticity—is what improves the metrics. the 4 major comments →

arxiv 2510.02208 v3 pith:M4W4NWJU submitted 2025-10-02 eess.IV cs.CVcs.LG

MACS: Measurement-Aware Consistency Sampling for Inverse Problems

Amirreza Tanevardi , Pooria Abbas Rad Moghadam , Seyed Mohammad Eshtehardian , Sajjad Amini , Babak Khalaj This is my paper
classification eess.IV cs.CVcs.LG
keywords inverse problemsconsistency modelsfew-step samplingmeasurement consistencyimage restorationtest-time adaptationsuper-resolutiondeblurring
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tries to establish that the measurement residual—how far a current estimate is from matching the observed data through the degradation operator—can stand in for the variance-compensation term in a consistency-model sampler. It introduces MACS, which scales the injected noise at each sampling step by this residual: larger residuals allow more exploration, smaller residuals refine. This is a purely test-time change that leaves the consistency model and any conditioning network untouched, and the authors show consistent perceptual and pixel-level gains across super-resolution, inpainting, and deblurring on two datasets using only two sampling steps. The derivation treats the residual as an upper bound on the variance term and absorbs the constant factors into a single tunable scalar gamma.

Core claim

The central claim is that the stochasticity of a consistency-based sampler can be regulated by the measurement residual instead of an unobserved ground-truth difference. Concretely, MACS replaces the variance term eta||x - x_hat||^2 in the adjusted DDIM update with gamma||y - A(x_hat)||^2, where A is the degradation operator. The paper argues that for linear operators this residual is bounded by a scaled version of the variance term plus a measurement-noise floor, and that with a suitable gamma the two are functionally interchangeable. This substitution produces an update that injects more noise when the estimate is far from the measurements and less when it is close, steering the few-step t

What carries the argument

The MACS update step is the central mechanism: given the model's clean prediction x_hat and residual direction eps_hat = x_t - x_hat, the next latent is x_s = x_hat + sqrt(rho + ((1 - sqrt(rho))/||eps_hat||)^2 * gamma*||y - A(x_hat)||^2) * eps_hat, where rho is the noise-level ratio and gamma is a hyperparameter. The residual term replaces the variance-compensation term from the adjusted DDIM sampler, modulating the injection scale as a function of measurement fidelity while keeping the update direction aligned with the model's own estimate. This single algebraic change is what allows measurement awareness to be injected into the sampling trajectory without retraining.

Load-bearing premise

The derivation only shows that the measurement residual is bounded above by the variance term times the operator norm plus a noise floor, and then treats that bound as justification for replacing the variance term with the residual; this requires a single scalar gamma to absorb the operator norm and noise floor across all images, noise levels, and even nonlinear operators.

What would settle it

Run a calibration experiment on a fixed linear operator A: set gamma from the derived bound (||A||_2^2 plus the measurement-noise contribution) and compare reconstruction quality against the paper's tuned gamma. If the tuned gamma deviates by more than a small factor from that value, or if the optimal gamma varies strongly across images or noise levels, the substitution is not fully consistent with the aDDIM variance-compensation mechanism.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the claim holds, any consistency-model inverse solver can adopt MACS as a plug-in sampling schedule with no retraining, since it only modifies the update step.
  • Two-step reconstruction becomes practical for real-time inverse imaging, cutting the cost by an order of magnitude relative to diffusion solvers that need hundreds of function evaluations.
  • The residual-guided noise injection tends to shift the distortion–perception trade-off toward better perceptual quality while keeping pixel-level metrics competitive, as the authors observe.
  • The same residual-guidance principle could extend to other few-step samplers that share the DDIM-style update structure, not just consistency models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The single scalar gamma bundles the operator norm, measurement noise variance, and the unknown conditional variance; a per-operator or per-step calibration of gamma based on ||A||_2 and sigma_y could turn the upper bound into a tighter equality and may improve robustness.
  • Because the residual term uses the forward operator directly, MACS could be adapted to non-Gaussian measurement noise by replacing the L2 residual with a log-likelihood term, a natural extension the paper does not explore.
  • The method's success on nonlinear deblurring suggests the residual behaves more like a general data-fidelity scheduler than a strict variance estimator; a theory that does not rely on linearity would be a valuable next step.
  • A concrete stress test would be to hold gamma fixed while varying the conditioning of A or the noise level; if the optimal gamma shifts widely, the claimed interchangeability with aDDIM's variance term is only approximate.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 3 minor

Summary. The paper proposes MACS, a test-time sampling strategy for consistency models in inverse problems. The core idea is to replace the variance-compensation term of the aDDIM sampler with a measurement residual γ||y−A(x̂)||² inside the noise-injection coefficient, so that the sampler's stochasticity is adjusted according to how well the current estimate matches the observations. The method is evaluated on Fashion-MNIST and LSUN Bedroom for super-resolution, inpainting, linear deblurring, and nonlinear deblurring, using fixed pretrained CM backbones and conditioning networks. The reported results show consistent FID/KID improvements at 2 NFE relative to ODE baselines (Euler, Heun, DPM-Solver) and standard multistep consistency sampling, with comparable or slightly lower PSNR/SSIM.

Significance. If the mechanism works as claimed, MACS would be a simple, retraining-free improvement to CM-based inverse solvers, and the empirical gains on two datasets and four tasks are suggestive. The paper's strengths are its clear experimental setup, comparison at matched NFE, and use of standard metrics. However, the theoretical justification in Eqs. (37)–(41) is only an upper bound, not a derivation of equivalence with aDDIM, and the experiments do not include an ablation that isolates measurement-awareness from added stochasticity. These gaps leave the central causal claim—that the measurement residual, rather than increased noise, is the active ingredient—unsupported. The contribution is thus best viewed at present as a promising heuristic rather than a validated mechanism.

major comments (4)
  1. [§IV, Eqs. (37)–(41)] The derivation establishes only E||y−A(x̂)||² ≤ ||A||₂²||x−x̂||² + mσ_y², and Eq. (41) absorbs ||A||₂² and mσ_y² into a tunable γ. This is an inequality, not an equivalence; it does not show that the residual is proportional to the conditional variance that aDDIM compensates. For rank-deficient operators (inpainting, super-resolution), the residual can be far smaller than Var[x|x_t]. The sentence following Eq. (41) that the residual is 'fully consistent' with aDDIM is therefore not justified. The theoretical link is load-bearing for the 'measurement-aware' claim.
  2. [§V-C, Eq. (36), Tables I–III] No ablation separates the effect of the residual's dependence on y from the effect of increased stochasticity. For any γ>0, the update (Eq. 36) adds more noise than DDIM/ODE baselines. Since the authors themselves cite the perception–distortion tradeoff, and extra stochasticity alone is known to improve FID/KID, the observed gains could arise simply from larger noise. A control with R replaced by a constant or by a y-independent variance proxy is needed to attribute the improvements to measurement-awareness. Without it, the central mechanism remains untested.
  3. [§V-B, Tables I–III] The guidance scale γ is chosen per task (0.15–0.8) with no sensitivity analysis, and the reported metrics appear to be point estimates without error bars or multiple seeds. For example, on LSUN SR the FID difference between MACS (40.13) and Multistep (40.85) is small; it is unclear if this exceeds run-to-run variability. Adding confidence intervals or a γ-sweep would strengthen the claim of consistent, robust gains.
  4. [§V-A, Table III] The theoretical derivation in Eqs. (37)–(41) assumes a linear forward operator A. The nonlinear deblurring experiment in Table III is presented as an extension of the same principle, but the spectral norm bound and the expectation expansion do not apply to nonlinear maps. The nonlinear result should be described as heuristic/empirical, or a separate argument must be provided; as written, the paper extends the theoretical justification beyond its domain of validity.
minor comments (3)
  1. [Table I] The Multistep row is difficult to read due to missing spacing (e.g., '226.03 0.77140.90 5.67' appears to combine the NFE '2' with PSNR '26.03' and SSIM '0.771'). Please reformat for clarity.
  2. [§II-D, Eq. (25)] The notation x_{var,t} is introduced as a scalar η||x−x̂||² but then appears multiplicatively with the vector ε̂. Please clarify whether it is a scalar coefficient or a vector, and how the product is computed.
  3. [§V-C] The claim that increasing NFE beyond 2 does not improve results is supported only by the residue curves in Fig. 4, not by metric tables at NFE=3 or 4. A small table or a sentence reporting metrics at higher NFE would make this observation reproducible.

Circularity Check

0 steps flagged

No significant circularity: MACS is an empirical sampler modification; its reported gains are measured, and its theoretical link to aDDIM is an approximation, not a derivation that assumes its own conclusion.

full rationale

The derivation chain does not reduce to its inputs. Eq. (36) defines the MACS update with the residual term γ||y−A(x̂)||²; this is a proposed rule, not a prediction derived from a model that already contains the result. The connection to aDDIM (Eqs. 37–41) is a heuristic bound: E||y−A(x̂)||² ≤ ||A||²₂||x−x̂||² + mσ²_y, followed by absorption into a tunable γ. This supports (weakly) a resemblance to the variance-compensation term η||x−x̂||², but it is not an identity and is not used to fit the reported metrics; it is an approximation/validation concern rather than a circular step. γ is selected per task empirically, but no fitted parameter is renamed as a prediction: the Tables report actual FID/KID/PSNR/SSIM against held-out baselines. There are no self-citations: the cited priors (CM [41], aDDIM [42], CoSIGN [54], SBI [55]) are independent prior work, and none of the authors' own prior results is load-bearing. The missing γ=0 / y-independent stochasticity ablation is a threat to the attribution of the gains to measurement-awareness, but that is an experimental-validity gap, not a circularity. Therefore the paper is not circular; score 0.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The central method introduces no new physical entities or learned networks. The free parameters are the per-task guidance scale γ and the number of steps; the main axiomatic burden is that a measurement residual can replace aDDIM's variance term based only on an upper bound, and that this replacement works for nonlinear operators.

free parameters (2)
  • γ (guidance scale) = LSUN: 0.8 (inpainting/SR), 0.4 (linear deblur), 0.5 (nonlinear deblur); Fashion-MNIST: 0.15
    Controls the magnitude of residual-based noise injection; selected empirically per task, with no sensitivity analysis or selection protocol reported.
  • Number of sampling steps N = 2
    Chosen empirically; the paper states increasing N beyond 2 does not improve metrics.
axioms (4)
  • ad hoc to paper aDDIM's variance-compensation framework is valid for CM few-step sampling and that replacing its teacher-signal variance with a measurement residual preserves the compensation.
    Used to justify Eq. (36) via the upper-bound argument in Eqs. (37)–(41); no equality or proportionality is established.
  • ad hoc to paper The measurement residual is approximately proportional to the conditional variance of x_t, absorbed into a single tunable scalar γ.
    The derivation only gives E||y−A(ˆx)||² ≤ ||A||₂²||x−ˆx||² + mσ_y², then absorbs constants into γ.
  • domain assumption The forward operator is linear and measurement noise is independent Gaussian with known variance for the theoretical derivation.
    Used in Eqs. (37)–(39); nevertheless applied to nonlinear deblurring in Table III without additional justification.
  • domain assumption The conditional consistency model f_θ(x_t, y, t) provides a reliable Tweedie-style clean estimate.
    Standard for CM/DM inverse solvers; not re-derived here.

pith-pipeline@v1.3.0-alltime-deepseek · 14022 in / 11854 out tokens · 93558 ms · 2026-08-04T12:41:49.341204+00:00 · methodology

0 comments
read the original abstract

Diffusion models have emerged as powerful generative priors for solving inverse imaging problems. However, their practical deployment is hindered by the substantial computational cost of slow, multi-step sampling. Although Consistency Models (CMs) address this limitation by enabling high-quality generation in only one or a few steps, their direct application to inverse problems has remained largely unexplored. This paper introduces a modified consistency sampling framework specifically designed for inverse problems. The proposed approach regulates the sampler's stochasticity through a measurement-consistency mechanism that leverages the degradation operator, thereby enforcing fidelity to the observed data while preserving the computational efficiency of consistency-based generation. Comprehensive experiments on the Fashion-MNIST and LSUN Bedroom datasets demonstrate consistent improvements across both perceptual and pixel-level metrics, including the Fr\'echet Inception Distance (FID), Kernel Inception Distance (KID), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM), compared with baseline consistency and diffusion-based sampling methods. The proposed method achieves competitive or superior reconstruction quality with only a small number of sampling steps.

Figures

Figures reproduced from arXiv: 2510.02208 by Amirreza Tanevardi, Babak Khalaj, Pooria Abbas Rad Moghadam, Sajjad Amini, Seyed Mohammad Eshtehardian.

Figure 1
Figure 1. Figure 1: Illustration of the proposed MACS framework. The diagram shows [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Visual comparison across four inverse problems on the LSUN Bedroom dataset: inpainting (top row), linear deblurring (second row), super-resolution [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visual comparison on three inverse problems from the Fashion-MNIST dataset: inpainting (top), linear deblurring (middle), and super-resolution [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Normalized measurement residue as a function of NFEs for MACS on the (a) LSUN Bedroom and (b) Fashion-MNIST datasets. MACS achieves [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

62 extracted references · 7 linked inside Pith

  1. [1]

    Solving inverse problems in medical imaging with score-based generative models,

    Y . Song, L. Shen, L. Xing, and S. Ermon, “Solving inverse problems in medical imaging with score-based generative models,” inProceedings of the International Conference on Learning Representations (ICLR), 2021

    2021

  2. [2]

    Compressed sensing us- ing generative models,

    A. Bora, A. Jalal, E. Price, and A. G. Dimakis, “Compressed sensing us- ing generative models,” inProceedings of the International Conference on Machine Learning (ICML), 2017, pp. 537–546

    2017

  3. [3]

    An overview of full-waveform inversion in exploration geophysics,

    J. Virieux and S. Operto, “An overview of full-waveform inversion in exploration geophysics,”Geophysics, vol. 74, no. 6, pp. WCC1–WCC26, 2009

    2009

  4. [4]

    Video diffusion models,

    J. Ho, T. Salimans, A. Gritsenko, W. Chan, M. Norouzi, and D. Fleet, “Video diffusion models,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022

    2022

  5. [5]

    Diffir2vr-zero: Zero-shot video restoration with diffusion- based image restoration models,

    C.-H. Yeh, C.-Y . Lin, Z. Wang, C.-W. Hsiao, T.-H. Chen, H.-S. Shiu, and Y .-L. Liu, “Diffir2vr-zero: Zero-shot video restoration with diffusion- based image restoration models,”arXiv preprint arXiv:2407.01519, 2024

    arXiv 2024

  6. [6]

    Warped diffusion: Solving video inverse problems with image diffusion models,

    G. Daras, W. Nie, K. Kreis, A. G. Dimakis, M. Mardani, N. Kovachki, and A. Vahdat, “Warped diffusion: Solving video inverse problems with image diffusion models,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 37, 2024

    2024

  7. [7]

    Vision-XL: High definition video inverse problem solver using latent image diffusion models,

    T. Kwon and J. C. Ye, “Vision-XL: High definition video inverse problem solver using latent image diffusion models,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025

    2025

  8. [8]

    Solving video inverse problems using image diffusion models,

    ——, “Solving video inverse problems using image diffusion models,” inProceedings of the International Conference on Learning Represen- tations (ICLR), 2025

    2025

  9. [9]

    Solving audio inverse prob- lems with a diffusion model,

    E. Moliner, J. Lehtinen, and V . V ¨alim¨aki, “Solving audio inverse prob- lems with a diffusion model,” inProceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023, pp. 1–5

    2023

  10. [10]

    Image denoising by sparse 3-D transform-domain collaborative filtering,

    K. Dabov, A. Foi, V . Katkovnik, and K. Egiazarian, “Image denoising by sparse 3-D transform-domain collaborative filtering,”IEEE Transactions on Image Processing, vol. 16, no. 8, pp. 2080–2095, 2007

    2080

  11. [11]

    Image denoising via sparse and redundant representations over learned dictionaries,

    M. Elad and M. Aharon, “Image denoising via sparse and redundant representations over learned dictionaries,”IEEE Transactions on Image Processing, vol. 15, no. 12, pp. 3736–3745, 2006

    2006

  12. [12]

    Weighted nuclear norm minimization with application to image denoising,

    S. Gu, L. Zhang, W. Zuo, and X. Feng, “Weighted nuclear norm minimization with application to image denoising,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014, pp. 2862–2869

    2014

  13. [13]

    Image denoising: The deep learning revolution and beyond—a survey paper,

    M. Elad, B. Kawar, and G. Vaksman, “Image denoising: The deep learning revolution and beyond—a survey paper,”SIAM Journal on Imaging Sciences, vol. 16, no. 3, pp. 1594–1654, 2023

    2023

  14. [14]

    A residual dense U-Net neural network for image denoising,

    J. Gurrola-Ramos, O. Dalmau, and T. E. Alarc ´on, “A residual dense U-Net neural network for image denoising,”IEEE Access, vol. 9, pp. 31 742–31 754, 2021

    2021

  15. [15]

    Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restoration,

    Y . Chen and T. Pock, “Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restoration,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1256– 1272, 2017

    2017

  16. [16]

    High perceptual quality image denoising with a posterior sampling cGAN,

    G. Ohayon, T. Adrai, G. Vaksman, M. Elad, and P. Milanfar, “High perceptual quality image denoising with a posterior sampling cGAN,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 1805–1813

    2021

  17. [17]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 33, 2020, pp. 6840–6851

    2020

  18. [18]

    Generative modeling by estimating gradients of the data distribution,

    Y . Song and S. Ermon, “Generative modeling by estimating gradients of the data distribution,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 32, 2019

    2019

  19. [19]

    Diffusion models beat GANs on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat GANs on image synthesis,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 34, 2021, pp. 8780–8794

    2021

  20. [20]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR), 2022, pp. 10 684–10 695

    2022

  21. [21]

    Score-based generative modeling through stochastic differ- ential equations,

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differ- ential equations,”arXiv preprint arXiv:2011.13456, 2020

    Pith/arXiv arXiv 2011

  22. [22]

    Denoising diffusion restoration models,

    B. Kawar, M. Elad, S. Ermon, and J. Song, “Denoising diffusion restoration models,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022, pp. 23 593–23 606

    2022

  23. [23]

    Deep back-projection networks for super-resolution,

    M. Haris, G. Shakhnarovich, and N. Ukita, “Deep back-projection networks for super-resolution,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 1664– 1673

    2018

  24. [24]

    Image super-resolution via iterative refinement,

    C. Saharia, J. Ho, W. Chan, T. Salimans, D. J. Fleet, and M. Norouzi, “Image super-resolution via iterative refinement,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 4, pp. 4713– 4726, 2023

    2023

  25. [25]

    Semantic image inpainting with deep generative models,

    R. A. Yeh, C. Chen, T. Y . Lim, A. G. Schwing, M. Hasegawa-Johnson, and M. N. Do, “Semantic image inpainting with deep generative models,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 5485–5493

    2017

  26. [26]

    RePaint: Inpainting using denoising diffusion probabilistic mod- els,

    A. Lugmayr, M. Danelljan, A. Romero, F. Yu, R. Timofte, and L. V . Gool, “RePaint: Inpainting using denoising diffusion probabilistic mod- els,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2022, pp. 11 461–11 471

    2022

  27. [27]

    DeblurGAN-v2: De- blurring (orders-of-magnitude) faster and better,

    O. Kupyn, T. Martyniuk, J. Wu, and Z. Wang, “DeblurGAN-v2: De- blurring (orders-of-magnitude) faster and better,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 8878–8887

    2019

  28. [28]

    Palette: Image-to-image diffusion models,

    C. Saharia, W. Chan, H. Chang, C. Lee, J. Ho, T. Salimans, D. Fleet, and M. Norouzi, “Palette: Image-to-image diffusion models,” inACM SIGGRAPH Conference Proceedings (SIGGRAPH), 2022

    2022

  29. [29]

    Solving inverse problems with latent diffusion models via hard data consistency,

    B. Song, S. M. Kwon, Z. Zhang, X. Hu, Q. Qu, and L. Shen, “Solving inverse problems with latent diffusion models via hard data consistency,” inProceedings of the International Conference on Learning Represen- tations (ICLR), 2024

    2024

  30. [30]

    SILO: Solving inverse problems with latent operators,

    R. Raphaeli, S. Man, and M. Elad, “SILO: Solving inverse problems with latent operators,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025, pp. 10 570–10 580

    2025

  31. [31]

    Improving diffusion models for inverse problems using manifold constraints,

    H. Chung, B. Sim, D. Ryu, and J. C. Ye, “Improving diffusion models for inverse problems using manifold constraints,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022, pp. 25 683– 25 696

    2022

  32. [32]

    Diffusion posterior sampling for general noisy inverse problems,

    H. Chung, J. Kim, M. T. McCann, M. L. Klasky, and J. C. Ye, “Diffusion posterior sampling for general noisy inverse problems,” inProceedings of the International Conference on Learning Representations (ICLR), 2023. 10

    2023

  33. [33]

    Decomposed diffusion sampler for accelerating large-scale inverse problems,

    H. Chung, S. Lee, and J. C. Ye, “Decomposed diffusion sampler for accelerating large-scale inverse problems,” inProceedings of the International Conference on Learning Representations (ICLR), 2024

    2024

  34. [34]

    Solving 3D inverse problems using pre-trained 2D diffusion models,

    H. Chung, D. Ryu, M. T. McCann, M. L. Klasky, and J. C. Ye, “Solving 3D inverse problems using pre-trained 2D diffusion models,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 22 542–22 551

    2023

  35. [35]

    SNIPS: Solving noisy inverse problems stochastically,

    B. Kawar, G. Vaksman, and M. Elad, “SNIPS: Solving noisy inverse problems stochastically,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 34, 2021, pp. 21 757–21 769

    2021

  36. [36]

    Pseudoinverse-guided diffusion models for inverse problems,

    J. Song, A. Vahdat, M. Mardani, and J. Kautz, “Pseudoinverse-guided diffusion models for inverse problems,” inProceedings of the Interna- tional Conference on Learning Representations (ICLR), 2023

    2023

  37. [37]

    Zero-shot image restoration using de- noising diffusion null-space model,

    Y . Wang, J. Yu, and J. Zhang, “Zero-shot image restoration using de- noising diffusion null-space model,” inProceedings of the International Conference on Learning Representations (ICLR), 2023

    2023

  38. [38]

    ILVR: Condition- ing method for denoising diffusion probabilistic models,

    J. H. Choi, S. Kim, Y . Jeong, Y . Gwon, and S. Yoon, “ILVR: Condition- ing method for denoising diffusion probabilistic models,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 14 367–14 376

    2021

  39. [39]

    Direct diffusion bridge using data consistency for inverse problems,

    H. Chung, J. Kim, and J. C. Ye, “Direct diffusion bridge using data consistency for inverse problems,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 36, 2023

    2023

  40. [40]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” inProceedings of the International Conference on Learning Represen- tations (ICLR), 2020

    2020

  41. [41]

    Consistency models,

    Y . Song, P. Dhariwal, M. Chen, and I. Sutskever, “Consistency models,” inProceedings of the International Conference on Machine Learning (ICML), vol. 202, 2023, pp. 32 211–32 252

    2023

  42. [42]

    Multistep consistency models,

    J. Heek, E. Hoogeboom, and T. Salimans, “Multistep consistency models,”arXiv preprint arXiv:2403.06807, 2024

    Pith/arXiv arXiv 2024

  43. [43]

    Consistency models made easy,

    Z. Geng, A. Pokle, W. Luo, J. Lin, and J. Z. Kolter, “Consistency models made easy,”arXiv preprint arXiv:2406.14548, 2024

    Pith/arXiv arXiv 2024

  44. [44]

    Fashion-MNIST: A novel image dataset for benchmarking machine learning algorithms,

    H. Xiao, K. Rasul, and R. V ollgraf, “Fashion-MNIST: A novel image dataset for benchmarking machine learning algorithms,”arXiv preprint arXiv:1708.07747, 2017

    Pith/arXiv arXiv 2017

  45. [45]

    LSUN: Construction of a large-scale image dataset using deep learning with humans in the loop,

    F. Yu, A. Seff, Y . Zhang, S. Song, T. Funkhouser, and J. Xiao, “LSUN: Construction of a large-scale image dataset using deep learning with humans in the loop,”arXiv preprint arXiv:1506.03365, 2015

    Pith/arXiv arXiv 2015

  46. [46]

    GANs trained by a two time-scale update rule converge to a local Nash equilibrium,

    M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochreiter, “GANs trained by a two time-scale update rule converge to a local Nash equilibrium,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017, pp. 6629–6640

    2017

  47. [47]

    Demysti- fying MMD GANs,

    M. Bi ´nkowski, D. J. Sutherland, M. Arbel, and A. Gretton, “Demysti- fying MMD GANs,” inProceedings of the International Conference on Learning Representations (ICLR), 2018

    2018

  48. [48]

    Image quality assessment: From error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: From error visibility to structural similarity,”IEEE Transactions on Image Processing, vol. 13, no. 4, pp. 600–612, 2004

    2004

  49. [49]

    Latent consistency models: Synthesizing high-resolution images with few-step inference,

    S. Luo, Y . Tan, L. Huang, J. Li, and H. Zhao, “Latent consistency models: Synthesizing high-resolution images with few-step inference,” arXiv preprint arXiv:2310.04378, 2023

    Pith/arXiv arXiv 2023

  50. [50]

    InvFusion: Bridging supervised and zero-shot diffusion for inverse problems,

    N. Elata, H. Chung, J. C. Ye, T. Michaeli, and M. Elad, “InvFusion: Bridging supervised and zero-shot diffusion for inverse problems,” in Advances in Neural Information Processing Systems (NeurIPS), 2025

    2025

  51. [51]

    Loss-guided diffusion models for plug-and-play con- trollable generation,

    J. Song, Q. Zhang, H. Yin, M. Mardani, M.-Y . Liu, J. Kautz, Y . Chen, and A. Vahdat, “Loss-guided diffusion models for plug-and-play con- trollable generation,” inProceedings of the International Conference on Machine Learning (ICML), 2023, pp. 32 483–32 498

    2023

  52. [52]

    DEFT: Efficient fine-tuning of diffusion models by learning the generalized h-transform,

    A. Denker, F. Vargas, S. Padhy, K. Didi, S. C. Mathis, R. Barbano, V . Dutordoir, E. Mathieu, U. J. Komorowska, and P. Li `o, “DEFT: Efficient fine-tuning of diffusion models by learning the generalized h-transform,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 37, 2024, pp. 19 636–19 682

    2024

  53. [53]

    A survey on diffusion models for inverse problems,

    G. Daras, H. Chung, C.-H. Lai, Y . Mitsufuji, J. C. Ye, P. Milanfar, A. G. Dimakis, and M. Delbracio, “A survey on diffusion models for inverse problems,”arXiv preprint arXiv:2410.00083, 2024

    Pith/arXiv arXiv 2024

  54. [54]

    CoSIGN: Few-step guidance of consistency model to solve general inverse problems,

    J. Zhao, B. Song, and L. Shen, “CoSIGN: Few-step guidance of consistency model to solve general inverse problems,” inProceedings of the European Conference on Computer Vision (ECCV), ser. Lecture Notes in Computer Science, vol. 13627, 2024, pp. 108–126

    2024

  55. [55]

    Con- sistency models for scalable and fast simulation-based inference,

    M. Schmitt, V . Pratz, U. K ¨othe, P.-C. B ¨urkner, and S. T. Radev, “Con- sistency models for scalable and fast simulation-based inference,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 37, 2024

    2024

  56. [56]

    LATINO-PRO: Latent consistency inverse solver with prompt opti- mization,

    A. Spagnoletti, J. Prost, A. Almansa, N. Papadakis, and M. Pereyra, “LATINO-PRO: Latent consistency inverse solver with prompt opti- mization,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025, pp. 19 597–19 607

    2025

  57. [57]

    Zero-shot image restoration using few-step guidance of consistency models (and beyond),

    T. Garber and T. Tirer, “Zero-shot image restoration using few-step guidance of consistency models (and beyond),” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025, pp. 2398–2407

    2025

  58. [58]

    DPM-Solver: A fast ODE solver for diffusion probabilistic model sampling in around 10 steps,

    C. Lu, Y . Zhou, F. Bao, J. Chen, C. Li, and J. Zhu, “DPM-Solver: A fast ODE solver for diffusion probabilistic model sampling in around 10 steps,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022

    2022

  59. [59]

    Elucidating the design space of diffusion-based generative models,

    T. Karras, M. Aittala, T. Aila, and S. Laine, “Elucidating the design space of diffusion-based generative models,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022

    2022

  60. [60]

    The perception-distortion tradeoff,

    Y . Blau and T. Michaeli, “The perception-distortion tradeoff,” inPro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  61. [61]

    A theory of the distortion- perception tradeoff in wasserstein space,

    D. Freirich, T. Michaeli, and R. Meir, “A theory of the distortion- perception tradeoff in wasserstein space,” inAdvances in Neural Infor- mation Processing Systems (NeurIPS 2021), 2021, pp. 25 661–25 672

    2021

  62. [62]

    Looks too good to be true: An information-theoretic analysis of hallucinations in generative restoration models,

    R. Cohen, I. Kligvasser, E. Rivlin, and D. Freedman, “Looks too good to be true: An information-theoretic analysis of hallucinations in generative restoration models,” inProceedings of the 38th Conference on Neural Information Processing Systems (NeurIPS 2024), 2024

    2024