Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Modular MeanFlow: Towards Stable and Scalable One-Step Generative Modeling

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a modular loss with annealed gradient flow enables stable, one-step generative training that reaches FID 3.41 on CIFAR-10.

desk verdict The central loss derivation is mis-scaled, and the experiments as written don't test the claimed method; the underlying scheduling idea is plausible but needs a corrected derivation and rerun. read the letter →

arxiv 2508.17426 v1 pith:KBXAFENM submitted 2025-08-24 cs.LG

Haochen You , Baojing Liu , Hongyang He This is my paper · ORCID
classification cs.LG
keywords one-stepgenerationmeanflowaveragevelocitygradientmodulationcurriculumlearningconsistencymodelsmatchinggenerativemodeling
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 argues that one-step generative modeling can be made both stable and expressive by learning a time-averaged velocity field with a tunable loss. Instead of committing to either full backpropagation or stop-gradient training, Modular MeanFlow anneals the gradient flow from coarse supervision to fully differentiable training. The claimed payoff is competitive sample quality, FID 3.41 on CIFAR-10 with a one-step error of 0.076, together with stable convergence and better generalization in low-data and out-of-distribution settings. If right, this gives a practical recipe for fast generation that avoids expensive higher-order derivatives.

What carries the argument

The carrying object is the modular loss $L_\lambda$ built on the stop-gradient modulation $SG_\lambda[z]=\lambda z+(1-\lambda)\text{stopgrad}(z)$, together with the warm-up schedule $\lambda(t_{\text{train}})=\min(1,t_{\text{train}}/T_{\text{warmup}})$. This interpolation controls how much of the Jacobian-vector product $\nabla_x u_\theta \cdot v$ is allowed to backpropagate: full gradient flow at $\lambda=1$, no gradient flow at $\lambda=0$, and a smooth schedule in between. The loss is derived from the MeanFlow identity so that it supplies a time-averaged velocity target, and the modulation is what makes training stable without sacrificing expressiveness.

What would settle it

Train the same architecture on CIFAR-10 for the same 800,000 steps with the target $(x_1-x_0)/(1-r)$ in place of $(x_1-x_0)/(t-r)$ while keeping the curriculum. If the average-velocity interpretation is correct, the corrected scaling should not hurt FID and should reduce one-step error; if the current scaling is the right one, the corrected loss should degrade.

Watch

Extended reading notes

Core claim

The central claim is that the MeanFlow differential identity, $v(x_t,t)=u(x_t,r,t)+(t-r)\frac{d}{dt}u(x_t,r,t)$, can be turned into a family of regression losses in which the network's average velocity and its local dynamics are compared with the displacement $(x_1-x_0)/(t-r)$, while gradient flow through the second-order term is controlled by a parameter $\lambda$. With $\lambda=0$ the loss is stop-gradient and stable; with $\lambda=1$ it is fully backpropagated and expressive; the curriculum $\lambda(t)=\min(1,t/T_{\text{warmup}})$ moves smoothly from one to the other. The paper claims this curriculum regime achieves the best balance, producing FID 3.41 and one-step error 0.076 on CIFAR-10, and that the framework subsumes consistency-model and flow-matching objectives as special cases.

Load-bearing premise

The derivation assumes that the finite difference $(x_1-x_0)/(t-r)$ equals the true average velocity over the interval, while for the linear interpolation used in training the true average is $(x_1-x_0)/(1-r)$; these coincide only at $r=0$.

Editorial extensions

If this is right

  • One-step sampling reduces to $x_0=x_1-u_\theta(x_1,0,1)$, so generation needs a single function evaluation.
  • The same training objective transfers beyond images to tasks such as ODE trajectory fitting and control-path synthesis.
  • The curriculum schedule gives a principled way to choose between stability and expressiveness, and it can be tuned per domain.
  • Because the framework unifies consistency-based and flow-matching losses, improvements to the curriculum or to the modulation should transfer to those methods.
  • The low-data and distribution-shift results suggest that one-step velocity learning can be applied where full diffusion training would overfit.

Reading between the lines

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

  • Beyond what the paper tests, the residual one-step error could be probed by replacing the target $(x_1-x_0)/(t-r)$ with the exact average velocity $(x_1-x_0)/(1-r)$ for the linear paths actually used in training.
  • The paper anneals a single scalar $\lambda$; an extension would be to anneal the velocity-matching term and the Jacobian term at different rates, which might further stabilize the later training phase.
  • If the loss family truly generalizes consistency models, then consistency-training techniques such as boundary conditioning or exponential moving averages could be imported into Modular MeanFlow to reduce one-step error further.
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 / 5 minor

Summary. The manuscript introduces Modular MeanFlow (MMF), a training objective for one-step generative modeling that combines a MeanFlow-inspired regression target with a stop-gradient interpolation parameter λ and a curriculum schedule over λ. The authors claim that this loss family unifies consistency and flow-matching objectives, avoids higher-order derivatives, and achieves FID 3.41 on CIFAR-10 with one-step sampling, with additional results on few-shot and out-of-distribution settings, ODE fitting, and control. The theoretical derivation starts from an identity linking instantaneous and average velocity and proposes the loss Lλ in Eq. (11); the empirical section compares λ=0, λ=0.5, λ=1, and curriculum variants against MeanFlow baselines.

Significance. The central idea of tuning gradient flow through a stop-gradient interpolation is practically relevant, and a well-supported method of this kind could be useful for one-step generation. However, as written the paper does not deliver a machine-checked proof, reproducible code, or a parameter-free derivation, and the main loss is derived from an inconsistent velocity surrogate. The reported FID and error numbers therefore cannot be attributed to the stated method. The empirical study also lacks error bars, a Consistency Model baseline, and any ImageNet-64 result promised in Section 5.1. If the derivation were corrected, the modular loss idea would still need a substantial experimental re-run; in its current form the manuscript does not establish its central claims.

major comments (3)
  1. [§4.2, Eq. (8)-(11); §5.1] The regression target is (x1-x0)/(t-r), but the average velocity of the training path defined in Section 5.1 is (x1-x0)/(1-r). Specifically, xt=(1-α)x0+αx1 with α=(t-r)/(1-r) gives x_t-x_r=((t-r)/(1-r))(x1-x0), hence u=(x_t-x_r)/(t-r)=(x1-x0)/(1-r). The target differs by the factor (1-r)/(t-r) and matches only at t=1. Since r<t is sampled with probability one, essentially all training examples are supervised against a mis-scaled vector. The total-derivative term in Eq. (11) does not repair this because u is constant along the linear path and its total derivative is zero. This invalidates the claimed MeanFlow objective and the attribution of the reported FID, one-step error, and generalization results.
  2. [§4.5, Table 1] The assertion that Lλ unifies and generalizes full MeanFlow, StopGrad MeanFlow, and Consistency Models is stated without derivation. Table 1 gives a qualitative comparison, but no formal statement or proof shows how specific choices of λ, r, and t recover those objectives. This matters because the claimed theoretical grounding rests on that unification.
  3. [§5.1, §5.2] The experimental section promises ImageNet-64 and the Consistency Model baseline in Section 5.1 but never reports results for either; no standard errors or confidence intervals are given for FID or one-step error; and no code or checkpoints are provided. Given that the main contribution is a training objective, the absence of these details makes the empirical superiority claims unverifiable.
minor comments (5)
  1. [§4.4, Eq. (12)] The schedule λ(ttrain) uses ttrain, but the time variable t is already used for the path in Eq. (2); please clarify the notation to avoid confusion.
  2. [§5.1] The interpolation parametrization with α=(t-r)/(1-r) makes x_r=x0 and x_t=x1, while the ODE in Eq. (1) is written with x1 as the prior and x0 as data; the direction of the path should be stated explicitly.
  3. [References] Several references appear unrelated to the statements they are attached to (e.g., [11] for the harmonic potential in Section 5.6, and [22]-[23] for consistency formulations in Section 1); please recheck.
  4. [§4.5] The claim that the JVP overhead is 'typically <20% extra time' is not accompanied by any measurement or protocol.
  5. [Figures 3 and 4] The bar charts and box plots give no error bars or sample sizes, and the text describing 'variance' in Figure 4b does not define what variance is plotted.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the MMF objective is a supervised regression against explicitly data-derived velocity targets, and the self-citations are not load-bearing. An internal scaling mismatch between Section 4.2 and Section 5.1 is a correctness risk, not a circularity.

full rationale

The central training objective in Eq. (11) regresses the network field u_theta against the explicit data-dependent target (x1-x0)/(t-r). This is a conventional supervised regression target, not a fitted parameter renamed as a prediction: the network is trained to reproduce a vector field computed from the sampled endpoints, and the reported FID / one-step-error numbers are external empirical evaluations of the trained model. The differential identity (Eq. 4) is derived directly from the definition of average velocity (Eq. 2) by the fundamental theorem of calculus; it is an input identity, not an imported uniqueness theorem. The paper's self-citations (refs. 11, 12, 42, 45) support only auxiliary remarks: the examples used for ODE/control tasks, the 'continuous spectrum' interpretation of lambda, and a conservation toolkit analogy. None of these citations supplies the load-bearing claim that L_lambda unifies consistency models and flow matching; that unification claim is asserted via Table 1 without an explicit reduction, which is an evidential gap rather than circularity. I also flag the internal inconsistency noted in Section 4.2 versus Section 5.1: for the training path xt = (1-alpha)x0 + alpha x1 with alpha = (t-r)/(1-r), the exact average velocity is (x1-x0)/(1-r), yet Eq. (8) uses (x1-x0)/(t-r). This means the implemented target is mis-scaled for most sampled intervals, so the reported results may not validate the stated MeanFlow-inspired objective. That is a correctness or validity problem, not a circularity problem, because the target is still externally data-derived rather than being equivalent to the network output by construction.

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

The central claim relies on a standard calculus identity plus several domain assumptions about the velocity field and interpolation. The main load-bearing assumption is that the instantaneous velocity can be replaced by the finite difference (x1-x0)/(t-r), which is not justified and is numerically wrong for the defined interpolation. No new physical entities are introduced.

free parameters (2)
  • warmup duration T_warmup = 100,000 steps
    Hand-chosen horizon for the curriculum schedule in Eq. (12); the paper does not justify this value or show sensitivity to it.
  • time-pair sampling distribution = not specified
    The paper says time pairs (r,t) are randomly sampled with 0 ≤ r < t ≤ 1 but does not specify the distribution; the loss behavior depends on it.
assumptions (5)
  • standard math The fundamental theorem of calculus and product rule applied to differentiate (t−r)u = ∫_r^t v dτ
    Used to derive Eq. (4) in Section 4.2.
  • domain assumption The ODE solution exists and the velocity field v is sufficiently smooth for the exchange of differentiation and integration
    Implicit in the MeanFlow framework, Section 3.
  • ad hoc to paper The average velocity u can be replaced by the instantaneous velocity v inside the total derivative in Eq. (8)
    This substitution is not an identity for finite intervals and is the core heuristic of the loss.
  • ad hoc to paper The finite difference (x1−x0)/(t−r) is a valid surrogate for the velocity in the gradient modulation term of Eq. (11)
    The paper uses this as the target direction without deriving it from the interpolation, and it is inconsistent with the defined interpolation for r>0.
  • domain assumption A linear interpolation between x0 and x1 defines valid training trajectories
    Standard in flow matching; stated in Section 5.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Modular MeanFlow: Towards Stable and Scalable One-Step Generative Modeling." pith.science (2026). https://pith.science/paper/KBXAFENM

@misc{pith2026250817426,
  author       = {Pith},
  title        = {Pith review of: Modular MeanFlow: Towards Stable and Scalable One-Step Generative Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KBXAFENM}},
  note         = {Machine review of arXiv:2508.17426}
}
read the original abstract

One-step generative modeling seeks to generate high-quality data samples in a single function evaluation, significantly improving efficiency over traditional diffusion or flow-based models. In this work, we introduce Modular MeanFlow (MMF), a flexible and theoretically grounded approach for learning time-averaged velocity fields. Our method derives a family of loss functions based on a differential identity linking instantaneous and average velocities, and incorporates a gradient modulation mechanism that enables stable training without sacrificing expressiveness. We further propose a curriculum-style warmup schedule to smoothly transition from coarse supervision to fully differentiable training. The MMF formulation unifies and generalizes existing consistency-based and flow-matching methods, while avoiding expensive higher-order derivatives. Empirical results across image synthesis and trajectory modeling tasks demonstrate that MMF achieves competitive sample quality, robust convergence, and strong generalization, particularly under low-data or out-of-distribution settings.

Figures

Figures reproduced from arXiv: 2508.17426 by the authors.

Figure 1
Figure 1. Effect of λ on training dynamics: curriculum smoothing improves convergence (a), while naive propagation leads to instability (b). 5.3 Ablation Study: Role of Stop-Gradient To better understand the effect of stop-gradient modulation, we conduct an ablation study by analyzing models trained with varying values of the gradient control parameter λ. This analysis helps reveal the trade-offs between stability, accuracy, … view at source ↗
Figure 2
Figure 2. Visualization of learned velocity vectors. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. FID comparison across training regimes (left) and test domains (right). [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Analysis of generation diversity and training stability under different training [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Analysis of interpolation consistency across models. Curriculum-based MMF [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: MMF applied to low-dimensional trajectory reconstruction tasks. In both ODE [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

Review history (2 revisions) →

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. CaloTrilogy: Toward a Breakthrough in One-Step, End-to-End, Physics-Guided Shower Generation for Modern Calorimeters

    hep-ex 2026-06 unverdicted novelty 6.0 of 10

    Presents CaloTrilogy, a unified one-step generative model for high-granularity calorimeter showers that combines velocity field integration, learned priors, and physics losses to match SOTA quality.

Reference graph

Works this paper leans on

49 extracted references · 20 canonical work pages · cited by 1 Pith paper

  1. [1]

    arXiv preprint arXiv:2310.08337 (2023)

    Bartosh, G., Vetrov, D., Naesseth, C.A.: Neural diffusion models. arXiv preprint arXiv:2310.08337 (2023)

    arXiv 2023

  2. [2]

    National Science Review11(12), nwae348 (2024)

    Chen, M., Mei, S., Fan, J., Wang, M.: Opportunities and challenges of diffusion models for generative ai. National Science Review11(12), nwae348 (2024)

    work page 2024

  3. [3]

    Advances in Neural Information Processing Systems36, 68552– 68575 (2023)

    Chen, S., Chewi, S., Lee, H., Li, Y., Lu, J., Salim, A.: The probability flow ode is provably fast. Advances in Neural Information Processing Systems36, 68552– 68575 (2023)

    work page 2023

  4. [4]

    In: The Thirteenth International Conference on Learning Representations (2025)

    Davtyan, A., Dadi, L.T., Cevher, V., Favaro, P.: Faster inference of flow-based gen- erative models via improved data-noise coupling. In: The Thirteenth International Conference on Learning Representations (2025)

    work page 2025

  5. [5]

    arXiv preprint arXiv:1410.8516 (2014)

    Dinh, L., Krueger, D., Bengio, Y.: Nice: Non-linear independent components esti- mation. arXiv preprint arXiv:1410.8516 (2014)

    arXiv 2014

  6. [6]

    arXiv preprint arXiv:1605.08803 (2016)

    Dinh, L., Sohl-Dickstein, J., Bengio, S.: Density estimation using real nvp. arXiv preprint arXiv:1605.08803 (2016)

    arXiv 2016

  7. [7]

    In: Proceedings of the Computer Vision and Pattern Recognition Con- ference

    Dong, L., Fan, Q., Guo, Y., Wang, Z., Zhang, Q., Chen, J., Luo, Y., Zou, C.: Tsd-sr: One-step diffusion with target score distillation for real-world image super- resolution. In: Proceedings of the Computer Vision and Pattern Recognition Con- ference. pp. 23174–23184 (2025)

    2025

  8. [8]

    In: Forty-first international conference on machine learning (2024)

    Esser, P., Kulal, S., Blattmann, A., Entezari, R., Müller, J., Saini, H., Levi, Y., Lorenz, D., Sauer, A., Boesel, F., et al.: Scaling rectified flow transformers for high-resolution image synthesis. In: Forty-first international conference on machine learning (2024)

    2024

Show all 49 references
  1. [9]

    arXiv preprint arXiv:2505.13447 (2025)

    Geng,Z.,Deng,M.,Bai,X.,Kolter,J.Z.,He,K.:Meanflowsforone-stepgenerative modeling. arXiv preprint arXiv:2505.13447 (2025)

    2025 arXiv

  2. [10]

    Advances in Neural Information Processing Systems36, 41914–41931 (2023)

    Geng, Z., Pokle, A., Kolter, J.Z.: One-step diffusion distillation via deep equilib- rium models. Advances in Neural Information Processing Systems36, 41914–41931 (2023)

    2023

  3. [11]

    He, H., Xie, H., Shen, G., Fu, B., You, H., Sanchez Silva, V.: 4s-classifier: empower- ing conservation through semi-supervised learning for rare and endangered species (2025) 14 H. You, B. Liu, and H. He

    2025

  4. [12]

    He, H., Xie, H., You, H., Sanchez Silva, V.: Semi-vim: bidirectional state space model for mitigating label imbalance in semi-supervised learning (2025)

    2025

  5. [13]

    Advances in neural information processing systems30 (2017)

    Heusel,M.,Ramsauer,H.,Unterthiner,T.,Nessler,B.,Hochreiter,S.:Ganstrained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems30 (2017)

    2017

  6. [14]

    Advances in neural information processing systems33, 6840–6851 (2020)

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020)

    2020

  7. [15]

    Advances in Neural Information Processing Systems37, 110464–110493 (2024)

    Holderrieth, P., Xu, Y., Jaakkola, T.: Hamiltonian score matching and genera- tive flows. Advances in Neural Information Processing Systems37, 110464–110493 (2024)

    2024

  8. [16]

    arXiv preprint arXiv:2408.12270 (2024)

    Jeha, P., Grathwohl, W., Andersen, M.R., Ek, C.H., Frellsen, J.: Variance reduc- tion of diffusion model’s gradients with taylor approximation-based control variate. arXiv preprint arXiv:2408.12270 (2024)

    2024 arXiv

  9. [17]

    arXiv preprint arXiv:2310.02279 (2023)

    Kim, D., Lai, C.H., Liao, W.H., Murata, N., Takida, Y., Uesaka, T., He, Y., Mitsu- fuji, Y., Ermon, S.: Consistency trajectory models: Learning probability flow ode trajectory of diffusion. arXiv preprint arXiv:2310.02279 (2023)

    2023 arXiv

  10. [18]

    arXiv preprint arXiv:2405.14822 (2024)

    Kim, D., Lai, C.H., Liao, W.H., Takida, Y., Murata, N., Uesaka, T., Mitsufuji, Y., Ermon, S.: Pagoda: Progressive growing of a one-step generator from a low- resolution diffusion teacher. arXiv preprint arXiv:2405.14822 (2024)

    2024 arXiv

  11. [19]

    Advances in neural information processing systems31 (2018)

    Kingma, D.P., Dhariwal, P.: Glow: Generative flow with invertible 1x1 convolu- tions. Advances in neural information processing systems31 (2018)

    2018

  12. [20]

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

    2009

  13. [21]

    Advances in neural information processing systems37, 63082–63109 (2024)

    Lee, S., Lin, Z., Fanti, G.: Improving the training of rectified flows. Advances in neural information processing systems37, 63082–63109 (2024)

    2024

  14. [22]

    arXiv preprint arXiv:2506.17290 (2025)

    Li, Y., Dong, J., Dong, Z., Yang, C., An, Z., Xu, Y.: Srkd: Towards efficient 3d point cloud segmentation via structure-and relation-aware knowledge distillation. arXiv preprint arXiv:2506.17290 (2025)

    2025 arXiv

  15. [23]

    arXiv preprint arXiv:2505.12079 (2025)

    Li, Y., Li, K., Yin, X., Yang, Z., Dong, J., Dong, Z., Yang, C., Tian, Y., Lu, Y.: Sepprune: Structured pruning for efficient deep speech separation. arXiv preprint arXiv:2505.12079 (2025)

    2025 arXiv

  16. [24]

    ICMR (2025)

    Li, Y., Long, Q., Zhou, Y., Zhang, R., Ning, Z., Zhu, Z., Zhou, Y., Wang, X., Xiao, M.: Comae: Comprehensive attribute exploration for zero-shot hashing. ICMR (2025)

    2025

  17. [25]

    arXiv preprint arXiv:2410.14720 (2024)

    Li, Y., Lu, Y., Dong, Z., Yang, C., Chen, Y., Gou, J.: Sglp: A similarity guided fast layer partition pruning for compressing large deep models. arXiv preprint arXiv:2410.14720 (2024)

    2024

  18. [26]

    arXiv:2507.02939 (2025)

    Li, Y., Yang, C., Zeng, H., Dong, Z., An, Z., Xu, Y., Tian, Y., Wu, H.: Frequency-aligned knowledge distillation for lightweight spatiotemporal forecast- ing. arXiv:2507.02939 (2025)

    2025 arXiv

  19. [27]

    arXiv preprint arXiv:2210.02747 (2022)

    Lipman, Y., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. arXiv preprint arXiv:2210.02747 (2022)

    2022 arXiv

  20. [28]

    Advances in Neural Information Processing Systems35, 5775–5787 (2022)

    Lu, C., Zhou, Y., Bao, F., Chen, J., Li, C., Zhu, J.: Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. Advances in Neural Information Processing Systems35, 5775–5787 (2022)

    2022

  21. [29]

    Advances in Neural Information Processing Systems 37, 115377–115408 (2024)

    Luo, W., Huang, Z., Geng, Z., Kolter, J.Z., Qi, G.j.: One-step diffusion distilla- tion through score implicit matching. Advances in Neural Information Processing Systems 37, 115377–115408 (2024)

    2024

  22. [30]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2025) Modular MeanFlow 15

    Ma, Z., Zhang, Y., Jia, G., Zhao, L., Ma, Y., Ma, M., Liu, G., Zhang, K., Ding, N., Li, J., et al.: Efficient diffusion models: A comprehensive survey from principles to practices. IEEE Transactions on Pattern Analysis and Machine Intelligence (2025) Modular MeanFlow 15

    2025

  23. [31]

    In: International conference on machine learning

    Pascanu, R., Mikolov, T., Bengio, Y.: On the difficulty of training recurrent neural networks. In: International conference on machine learning. pp. 1310–1318. Pmlr (2013)

    2013

  24. [32]

    Applied Vegetation Science28(2), e70022 (2025)

    Rátiva Gaona, D., Wehncke, E.V., Mariano, N.A., Osorio Beristain, M.: Hydro- chory, a key ecological function of a tropical dry forest river threatened by a dam and open-pit coal mining in colombia. Applied Vegetation Science28(2), e70022 (2025)

    2025

  25. [33]

    arXiv preprint arXiv:2202.00512 (2022)

    Salimans, T., Ho, J.: Progressive distillation for fast sampling of diffusion models. arXiv preprint arXiv:2202.00512 (2022)

    2022 arXiv

  26. [34]

    arXiv preprint arXiv:2310.14189 (2023)

    Song, Y., Dhariwal, P.: Improved techniques for training consistency models. arXiv preprint arXiv:2310.14189 (2023)

    2023 arXiv

  27. [35]

    Song, Y., Dhariwal, P., Chen, M., Sutskever, I.: Consistency models (2023)

    2023

  28. [36]

    arXiv preprint arXiv:2011.13456 (2020)

    Song, Y., Sohl-Dickstein, J., Kingma, D.P., Kumar, A., Ermon, S., Poole, B.: Score- based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456 (2020)

    2020 arXiv

  29. [37]

    arXiv preprint arXiv:2410.18958 (2024)

    Wang, F.Y., Geng, Z., Li, H.: Stable consistency tuning: Understanding and im- proving consistency models. arXiv preprint arXiv:2410.18958 (2024)

    2024 arXiv

  30. [38]

    arXiv preprint arXiv:2504.20179 (2025)

    Wang, J., Zhang, D., Luo, J., Yang, Y., Luo, F.: Integration flow models. arXiv preprint arXiv:2504.20179 (2025)

    2025 arXiv

  31. [39]

    arXiv preprint arXiv:2505.20755 (2025)

    Wang, Y., Bai, W., Zhang, C., Zhang, D., Luo, W., Sun, H.: Uni-instruct: One-step diffusion model through unified diffusion divergence instruction. arXiv preprint arXiv:2505.20755 (2025)

    2025

  32. [40]

    Advances in Neural Information Processing Systems37, 45073–45104 (2024)

    Xie, S., Xiao, Z., Kingma, D., Hou, T., Wu, Y.N., Murphy, K.P., Salimans, T., Poole, B., Gao, R.: Em distillation for one-step diffusion models. Advances in Neural Information Processing Systems37, 45073–45104 (2024)

    2024

  33. [41]

    arXiv preprint arXiv:2502.13394 (2025)

    Xie, Y., Cheng, X.: Flow-based generative models as iterative algorithms in prob- ability space. arXiv preprint arXiv:2502.13394 (2025)

    2025 arXiv

  34. [42]

    Xu,S.,Ye,Y.,Li,M.,You,H.,Wang,K.,Zhang,W.:Drco:atoolkitforintelligently curbing illegal wildlife trade (2025)

    2025

  35. [43]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Yin, T., Gharbi, M., Zhang, R., Shechtman, E., Durand, F., Freeman, W.T., Park, T.: One-step diffusion with distribution matching distillation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 6613– 6623 (2024)

    2024

  36. [44]

    In: International Con- ference on Neural Information Processing

    You, H., Liu, B.: Application of pseudometric functions in clustering and a novel similarity measure based on path information discrepancy. In: International Con- ference on Neural Information Processing. pp. 59–73. Springer (2024)

    2024

  37. [45]

    arXiv preprint arXiv:2508.12149 (2025)

    You, H., Liu, B.: Mover: Multimodal optimal transport with volume-based embed- ding regularization. arXiv preprint arXiv:2508.12149 (2025)

    2025 arXiv

  38. [46]

    Knowledge-Based Systems (2025)

    Zeng, H., Li, Y., Niu, R., Yang, C., Wen, S.: Enhancing spatiotemporal prediction through the integration of mamba state space models and diffusion transformers. Knowledge-Based Systems (2025)

    2025

  39. [47]

    Advances in Neural Information Processing Systems 33, 15511–15521 (2020)

    Zhang, B., Jin, J., Fang, C., Wang, L.: Improved analysis of clipping algorithms for non-convex optimization. Advances in Neural Information Processing Systems 33, 15511–15521 (2020)

    2020

  40. [48]

    In: International conference on machine learning

    Zheng, H., Nie, W., Vahdat, A., Azizzadenesheli, K., Anandkumar, A.: Fast sam- pling of diffusion models via operator learning. In: International conference on machine learning. pp. 42390–42402. PMLR (2023)

    2023

  41. [49]

    arXiv e-prints pp

    Zhengyu Huang, D., Huang, J., Lin, Z.: Convergence analysis of probability flow ode for score-based generative models. arXiv e-prints pp. arXiv–2404 (2024)

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.