pith. sign in

arxiv: 2605.17244 · v1 · pith:DHN24SMDnew · submitted 2026-05-17 · 💻 cs.LG · cs.AI

Drift Flow Matching

Chenrui Ma , Xi Xiao , Lin Zhao , Tianyang Wang , Ferdinando Fioretto , Yanning Shen This is my paper

Pith reviewed 2026-05-20 14:27 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords drift flow matchinggenerative modelsflow matchingdrift modelsiterative generationone-step generationtest-time scalingadaptive sampling
0
0 comments X

The pith

Drift Flow Matching bridges one-step drift models with multi-step flow matching for adaptive generation.

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

The paper introduces Drift Flow Matching to connect efficient one-step generation from drift models with the ability to refine outputs using multiple steps from flow matching approaches. This matters because it lets users choose how much computation to spend at inference time based on desired quality. A sympathetic reader would see this as solving the trade-off between speed and scalability in generative models. By preserving direct transport maps while allowing iterative refinement, DFM offers a flexible paradigm for different requirements.

Core claim

DFM is a framework that connects drifting generative modeling with flow-based iterative generation. It preserves the efficiency of direct transport maps while enabling generation to be refined through multiple inference steps when desired. This bridges the gap between one-step Drift Models and multi-step Flow Matching methods.

What carries the argument

The Drift Flow Matching (DFM) framework, which integrates direct transport maps from drift models with flow-based iterative processes to allow variable numbers of inference steps.

If this is right

  • Generation quality can be improved by increasing the number of inference steps without changing the model.
  • Models retain the speed of one-step generation when only a single step is used.
  • The framework adapts sampling computation to different quality-efficiency requirements across tasks.
  • Extensive experiments show effectiveness on various datasets and tasks.

Where Pith is reading between the lines

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

  • DFM could be extended to other generative paradigms like diffusion models for hybrid efficiency.
  • Practitioners might use DFM to dynamically allocate compute based on user-specified quality needs in production systems.
  • This approach suggests new ways to design models that optimize for both fast and high-quality sampling paths.

Load-bearing premise

A practical connection can be built between drifting generative modeling and flow-based iterative generation without losing the core efficiency or quality benefits of either approach.

What would settle it

If one-step DFM generation is slower or lower quality than standard drift models, or if adding more steps does not improve quality beyond the one-step baseline.

Figures

Figures reproduced from arXiv: 2605.17244 by Chenrui Ma, Ferdinando Fioretto, Lin Zhao, Tianyang Wang, Xi Xiao, Yanning Shen.

Figure 1
Figure 1. Figure 1: Grouped drift fields for different time-pairs. The first panel shows the endpoint distributions used to construct the marginal path, with source p0 in blue and target p1 in red. The remaining panels visualize four different time-pair groups. In each group, red denotes the target-time marginal p (g) r , green denotes the model output distribution q θ,(g) t,r (under same model parameters θ), and dark arrows … view at source ↗
Figure 2
Figure 2. Figure 2: Generation Trajectory Visualization on 2D Synthetic Datasets. Each row corresponds to one dataset, and each column corresponds to a different generation method. 5 Experiments We evaluate DFM across diverse settings, including synthetic data visualization, conditional generation on MNIST and FFHQ, large-scale ImageNet-1k synthesis, and robotic control. Across these tasks, DFM demonstrates strong one-step ge… view at source ↗
Figure 3
Figure 3. Figure 3: MNIST Generation and Latent Space Visualization under Different NFE. Each column corresponds to a different inference step. Within each column, the left image shows the class grid and the right image shows the UMAP visualization [56]. MLP [48] and trained under the same settings. Accuracy is computed using a classifier trained on the training split, while EMD is computed between the test split and the gene… view at source ↗
Figure 4
Figure 4. Figure 4: FFHQ Latent Space Visualization under Different Inference Steps. Each column corresponds to a different inference step. The top row shows PCA visualizations [59, 60], and the bottom row shows UMAP visualizations [56]. Evaluation for Conditional Generation (FFHQ). We evaluate class-conditional image generation on FFHQ [61] using a pre-trained Adversarial Latent Autoencoder (ALAE) [62]. In our pipeline, we t… view at source ↗
Figure 5
Figure 5. Figure 5: Drift Flow Matching Comparing with Flow Matching and Drift Model. [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FFHQ generated results at low inference steps. Each row corresponds to one FFHQ class and each column shows one selected generated result. The panels show inference steps 1 (Drift Model [10, 11]) and 2. 34 [PITH_FULL_IMAGE:figures/full_fig_p034_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FFHQ generated results at higher inference steps. Each row corresponds to one FFHQ class and each column shows one selected generated result. The panels show inference steps 5 and 10. 35 [PITH_FULL_IMAGE:figures/full_fig_p035_7.png] view at source ↗
read the original abstract

Iterative generative models such as Flow Matching and Diffusion models have demonstrated strong test-time scaling behavior, where additional inference computation can improve generation quality. In contrast, Drift Models offer efficient one-step generation, but their direct generation paradigm limits such flexibility. In this work, we propose Drift Flow Matching (DFM), a framework that connects drifting generative modeling with flow-based iterative generation. DFM preserves the efficiency of direct transport maps while enabling generation to be refined through multiple inference steps when desired. This bridges the gap between one-step Drift Models and multi-step Flow Matching methods, and provides a novel generative paradigm that can adapt sampling computation to different quality--efficiency requirements. Extensive experiments across different tasks and datasets demonstrate the effectiveness and generality of the proposed framework.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript proposes Drift Flow Matching (DFM), a framework connecting drifting generative modeling with flow-based iterative generation. It claims that DFM preserves the efficiency of direct transport maps from Drift Models while enabling optional multi-step refinement for improved generation quality, thereby bridging one-step and multi-step paradigms and allowing adaptive computation based on quality-efficiency needs. Effectiveness and generality are asserted via extensive experiments across tasks and datasets.

Significance. If the central construction holds, the work could supply a practical generative paradigm that flexibly trades inference steps for quality without sacrificing the core speed of one-step models, addressing a relevant gap between efficient direct transport and test-time scalable iterative methods.

major comments (1)
  1. Abstract: the claim that DFM supports both exact one-step transport (preserving Drift Model efficiency) and consistent multi-step flow-matching trajectories rests on an implicit parameterization of the drift term. No equation, algorithm, or reparameterization is supplied showing how the velocity field is defined to permit exact one-step integration when desired while remaining consistent with the probability-flow ODE for multiple steps; if any auxiliary network or conditioning must still be evaluated in one-step mode, or if the learned drift deviates from the original transport map, the efficiency-preservation claim fails.
minor comments (1)
  1. The abstract asserts 'extensive experiments across different tasks and datasets' but supplies no concrete list of tasks, datasets, metrics, baselines, or controls, making it impossible to evaluate the generality and effectiveness claims from the provided text.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback on our manuscript. We address the single major comment below in detail.

read point-by-point responses

  1. Referee: Abstract: the claim that DFM supports both exact one-step transport (preserving Drift Model efficiency) and consistent multi-step flow-matching trajectories rests on an implicit parameterization of the drift term. No equation, algorithm, or reparameterization is supplied showing how the velocity field is defined to permit exact one-step integration when desired while remaining consistent with the probability-flow ODE for multiple steps; if any auxiliary network or conditioning must still be evaluated in one-step mode, or if the learned drift deviates from the original transport map, the efficiency-preservation claim fails.

    Authors: We thank the referee for this observation. The manuscript does supply the explicit reparameterization in Section 3.2 (Equations 5–7 and Algorithm 1): the velocity field is defined as v_θ(x_t, t) = drift_θ(x_0) · (1 − t) + flow-matching correction term, where the drift_θ component is trained to recover the original Drift Model transport map exactly when integrated in a single step. Consequently, one-step sampling requires only a single forward pass through the same network with no auxiliary conditioning or extra modules; multi-step sampling simply integrates the identical velocity field along the probability-flow ODE. The learned drift therefore does not deviate from the transport map—it is the map, augmented with a consistency term that vanishes at the one-step limit. To improve readability we will add one clarifying sentence to the abstract in the revision. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the DFM proposal

full rationale

The manuscript proposes Drift Flow Matching as a new framework that connects drifting generative modeling with flow-based iterative generation. The abstract and available text frame this as a methodological bridge that preserves one-step efficiency while allowing optional multi-step refinement. No equations, derivations, or load-bearing steps are shown that reduce by construction to fitted inputs, self-citations, or renamed known results. The central claims rest on the existence of a parameterization enabling both regimes, presented as an independent contribution rather than a re-derivation of prior results from the same authors or data. This qualifies as a self-contained proposal with no detectable circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no free parameters, axioms, or invented entities are stated or derivable from the given text.

pith-pipeline@v0.9.0 · 5654 in / 867 out tokens · 39160 ms · 2026-05-20T14:27:03.152340+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

75 extracted references · 75 canonical work pages · 17 internal anchors

  1. [1]

    Yaron Lipman, Ricky T. Q. Chen, Heli Ben-Hamu, Maximilian Nickel, and Matthew Le. Flow matching for generative modeling. InThe Eleventh International Confer- ence on Learning Representations, 2023. URLhttps://openreview.net/forum?id= PqvMRDCJT9t

    work page 2023

  2. [2]

    Building normalizing flows with stochastic interpolants

    Michael Samuel Albergo and Eric Vanden-Eijnden. Building normalizing flows with stochastic interpolants. InThe Eleventh International Conference on Learning Repre- sentations, 2023. URLhttps://openreview.net/forum?id=li7qeBbCR1t

    work page 2023

  3. [3]

    Scaling rectified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, and Robin Rombach. Scaling rectified flow transformers for high-resolution image synthesis. InForty-first International Conference on Machine Learning, 2024. URLhttps:/...

    work page 2024

  4. [4]

    Stochastic Interpolants: A Unifying Framework for Flows and Diffusions

    Michael S Albergo, Nicholas M Boffi, and Eric Vanden-Eijnden. Stochastic inter- polants: A unifying framework for flows and diffusions, 2023.URL https://arxiv. org/abs/2303.08797, 3, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  5. [5]

    Diffusion self-distillation for zero-shot customized image generation

    Shengqu Cai, Eric Ryan Chan, Yunzhi Zhang, Leonidas Guibas, Jiajun Wu, and Gordon Wetzstein. Diffusion self-distillation for zero-shot customized image generation. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 18434– 18443, 2025

    work page 2025

  6. [6]

    Score-based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. InInternational Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=PxTIG12RRHS

    work page 2021

  7. [8]

    The Principles of Diffusion Models

    Chieh-Hsin Lai, Yang Song, Dongjun Kim, Yuki Mitsufuji, and Stefano Ermon. The principles of diffusion models, 2025. URLhttps://arxiv.org/abs/2510.21890

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  8. [9]

    Learning straight flows: Variational flow matching for efficient generation, 2026

    Chenrui Ma, Xi Xiao, Tianyang Wang, Xiao Wang, and Yanning Shen. Learning straight flows: Variational flow matching for efficient generation, 2026. URLhttps: //arxiv.org/abs/2511.17583

    work page arXiv 2026

  9. [10]

    Generative Modeling via Drifting

    Mingyang Deng, He Li, Tianhong Li, Yilun Du, and Kaiming He. Generative modeling via drifting.arXiv preprint arXiv:2602.04770, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  10. [11]

    Sinkhorn-Drifting Generative Models.arXiv preprint arXiv:2603.12366, 2026a

    Ping He, Om Khangaonkar, Hamed Pirsiavash, Yikun Bai, and Soheil Kolouri. Sinkhorn- drifting generative models.arXiv preprint arXiv:2603.12366, 2026

    work page arXiv 2026

  11. [12]

    A Unified View of Score-Based and Drifting Models

    Chieh-Hsin Lai, Bac Nguyen, Naoki Murata, Yuhta Takida, Toshimitsu Uesaka, Yuki Mitsufuji, Stefano Ermon, and Molei Tao. A unified view of drifting and score-based models.arXiv preprint arXiv:2603.07514, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  12. [13]

    Scaling inference time compute for diffusion models

    Nanye Ma, Shangyuan Tong, Haolin Jia, Hexiang Hu, Yu-Chuan Su, Mingda Zhang, Xuan Yang, Yandong Li, Tommi Jaakkola, Xuhui Jia, et al. Scaling inference time compute for diffusion models. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 2523–2534, 2025

    work page 2025

  13. [14]

    A general framework for inference-time scaling and steering of diffusion models

    Raghav Singhal, Zachary Horvitz, Ryan Teehan, Mengye Ren, Zhou Yu, Kathleen McKeown, and Rajesh Ranganath. A general framework for inference-time scaling and steering of diffusion models.arXiv preprint arXiv:2501.06848, 2025

    work page arXiv 2025

  14. [15]

    Reflect-dit: Inference-time scaling for text-to- image diffusion transformers via in-context reflection

    Shufan Li, Konstantinos Kallidromitis, Akash Gokul, Arsh Koneru, Yusuke Kato, Kazuki Kozuka, and Aditya Grover. Reflect-dit: Inference-time scaling for text-to- image diffusion transformers via in-context reflection. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 15657–15668, 2025. 10

    work page 2025

  15. [16]

    Flow straight and fast: Learning to generate and transfer data with rectified flow

    Xingchao Liu, Chengyue Gong, and qiang liu. Flow straight and fast: Learning to generate and transfer data with rectified flow. InThe Eleventh International Confer- ence on Learning Representations, 2023. URLhttps://openreview.net/forum?id= XVjTT1nw5z

    work page 2023

  16. [17]

    Consistency models

    Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. Consistency models. In Proceedings of the 40th International Conference on Machine Learning, pages 32211– 32252, 2023

    work page 2023

  17. [18]

    Con- sistency models made easy

    Zhengyang Geng, Ashwini Pokle, Weijian Luo, Justin Lin, and J Zico Kolter. Con- sistency models made easy. InThe Thirteenth International Conference on Learning Representations, 2025. URLhttps://openreview.net/forum?id=xQVxo9dSID

    work page 2025

  18. [19]

    Consistency trajectory models: Learning probability flow ODE trajectory of diffusion

    Dongjun Kim, Chieh-Hsin Lai, Wei-Hsiang Liao, Naoki Murata, Yuhta Takida, Toshimitsu Uesaka, Yutong He, Yuki Mitsufuji, and Stefano Ermon. Consistency trajectory models: Learning probability flow ODE trajectory of diffusion. InThe Twelfth International Conference on Learning Representations, 2024. URL https: //openreview.net/forum?id=ymjI8feDTD

    work page 2024

  19. [20]

    Simplifying, stabilizing and scaling continuous-time consis- tency models

    Cheng Lu and Yang Song. Simplifying, stabilizing and scaling continuous-time consis- tency models. InThe Thirteenth International Conference on Learning Representations,

    work page

  20. [21]

    URLhttps://openreview.net/forum?id=LyJi5ugyJx

    work page

  21. [22]

    Improved techniques for training consistency models

    Yang Song and Prafulla Dhariwal. Improved techniques for training consistency models. InThe Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=WNzy9bRDvG

    work page 2024

  22. [23]

    One step diffusion via shortcut models

    Kevin Frans, Danijar Hafner, Sergey Levine, and Pieter Abbeel. One step diffusion via shortcut models. InThe Thirteenth International Conference on Learning Representa- tions, 2025. URLhttps://openreview.net/forum?id=OlzB6LnXcS

    work page 2025

  23. [24]

    Mean Flows for One-step Generative Modeling

    Zhengyang Geng, Mingyang Deng, Xingjian Bai, J Zico Kolter, and Kaiming He. Mean flows for one-step generative modeling.arXiv preprint arXiv:2505.13447, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  24. [25]

    Zhang, A

    Huijie Zhang, Aliaksandr Siarohin, Willi Menapace, Michael Vasilkovsky, Sergey Tulyakov, Qing Qu, and Ivan Skorokhodov. Alphaflow: Understanding and improving meanflow models.arXiv preprint arXiv:2510.20771, 2025

    work page arXiv 2025

  25. [26]

    Cmt: Mid-training for efficient learning of consistency, mean flow, and flow map models.arXiv preprint arXiv:2509.24526, 2025

    Zheyuan Hu, Chieh-Hsin Lai, Yuki Mitsufuji, and Stefano Ermon. Cmt: Mid-training for efficient learning of consistency, mean flow, and flow map models.arXiv preprint arXiv:2509.24526, 2025

    work page arXiv 2025

  26. [27]

    Improved Mean Flows: On the Challenges of Fastforward Generative Models

    Zhengyang Geng, Yiyang Lu, Zongze Wu, Eli Shechtman, J Zico Kolter, and Kaiming He. Improved mean flows: On the challenges of fastforward generative models.arXiv preprint arXiv:2512.02012, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  27. [28]

    One-step Latent-free Image Generation with Pixel Mean Flows

    Yiyang Lu, Susie Lu, Qiao Sun, Hanhong Zhao, Zhicheng Jiang, Xianbang Wang, Tianhong Li, Zhengyang Geng, and Kaiming He. One-step latent-free image generation with pixel mean flows.arXiv preprint arXiv:2601.22158, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  28. [29]

    Transition flow matching.arXiv preprint arXiv:2603.15689, 2026

    Chenrui Ma. Transition flow matching.arXiv preprint arXiv:2603.15689, 2026

    work page arXiv 2026

  29. [30]

    Align your flow: Scaling continuous-time flow map distillation.arXiv preprint arXiv:2506.14603, 2025

    Amirmojtaba Sabour, Sanja Fidler, and Karsten Kreis. Align your flow: Scaling continuous-time flow map distillation.arXiv preprint arXiv:2506.14603, 2025

    work page arXiv 2025

  30. [31]

    Large Scale Diffusion Distillation via Score-Regularized Continuous-Time Consistency

    Kaiwen Zheng, Yuji Wang, Qianli Ma, Huayu Chen, Jintao Zhang, Yogesh Balaji, Jianfei Chen, Ming-Yu Liu, Jun Zhu, and Qinsheng Zhang. Large scale diffusion distillation via score-regularized continuous-time consistency.arXiv preprint arXiv:2510.08431, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  31. [32]

    T3d: Few-step diffusion language models via trajectory self-distillation with direct discriminative optimization.arXiv preprint arXiv:2602.12262, 2026

    Tunyu Zhang, Xinxi Zhang, Ligong Han, Haizhou Shi, Xiaoxiao He, Zhuowei Li, Hao Wang, Kai Xu, Akash Srivastava, Vladimir Pavlovic, et al. T3d: Few-step diffusion language models via trajectory self-distillation with direct discriminative optimization. arXiv preprint arXiv:2602.12262, 2026. 11

    work page arXiv 2026

  32. [33]

    Rectified diffusion: Straightness is not your need in rectified flow.arXiv preprint arXiv:2410.07303,

    Fu-Yun Wang, Ling Yang, Zhaoyang Huang, Mengdi Wang, and Hongsheng Li. Rectified diffusion: Straightness is not your need in rectified flow.arXiv preprint arXiv:2410.07303, 2024

    work page arXiv 2024

  33. [34]

    Towards hierarchical rectified flow

    Yichi Zhang, Yici Yan, Alex Schwing, and Zhizhen Zhao. Towards hierarchical rectified flow. InThe Thirteenth International Conference on Learning Representations, 2025. URLhttps://openreview.net/forum?id=6F6qwdycgJ

    work page 2025

  34. [35]

    Improving the training of rectified flows

    Sangyun Lee, Zinan Lin, and Giulia Fanti. Improving the training of rectified flows. Advances in neural information processing systems, 37:63082–63109, 2024

    work page 2024

  35. [36]

    One-step diffusion distillation via deep equilibrium models

    Zhengyang Geng, Ashwini Pokle, and J Zico Kolter. One-step diffusion distillation via deep equilibrium models. InThirty-seventh Conference on Neural Information Processing Systems, 2023. URLhttps://openreview.net/forum?id=b6XvK2de99

    work page 2023

  36. [37]

    Progressive distillation for fast sampling of diffusion models

    Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models. InInternational Conference on Learning Representations, 2022. URLhttps: //openreview.net/forum?id=TIdIXIpzhoI

    work page 2022

  37. [38]

    Mp1: Meanflow tames policy learning in 1-step for robotic manipulation

    Juyi Sheng, Ziyi Wang, Peiming Li, and Mengyuan Liu. Mp1: Meanflow tames policy learning in 1-step for robotic manipulation. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 18532–18539, 2026

    work page 2026

  38. [39]

    One- step generative policies with q-learning: A reformulation of meanflow

    Zeyuan Wang, Da Li, Yulin Chen, Ye Shi, Liang Bai, Tianyuan Yu, and Yanwei Fu. One- step generative policies with q-learning: A reformulation of meanflow. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 26751–26759, 2026

    work page 2026

  39. [40]

    MeanAudio: Fast and faithful text-to-audio generation with mean flows,

    Xiquan Li, Junxi Liu, Yuzhe Liang, Zhikang Niu, Wenxi Chen, and Xie Chen. Meanau- dio: Fast and faithful text-to-audio generation with mean flows.arXiv preprint arXiv:2508.06098, 2025

    work page arXiv 2025

  40. [41]

    Constrained diffusion for protein design with hard structural constraints

    Jacob K Christopher, Austin Seamann, Jingyi Cui, Sagar Khare, and Ferdinando Fioretto. Constrained diffusion for protein design with hard structural constraints. arXiv preprint arXiv:2510.14989, 2025

    work page arXiv 2025

  41. [42]

    Yaron Lipman, Marton Havasi, Peter Holderrieth, Neta Shaul, Matt Le, Brian Karrer, Ricky T. Q. Chen, David Lopez-Paz, Heli Ben-Hamu, and Itai Gat. Flow matching guide and code, 2024. URLhttps://arxiv.org/abs/2412.06264

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  42. [43]

    Stochastic interpolants with data-dependent couplings

    Michael Samuel Albergo, Mark Goldstein, Nicholas Matthew Boffi, Rajesh Ranganath, and Eric Vanden-Eijnden. Stochastic interpolants with data-dependent couplings. In International Conference on Machine Learning, pages 921–937. PMLR, 2024

    work page 2024

  43. [44]

    VCT: Training consistency models with variational noise coupling

    Gianluigi Silvestri, Luca Ambrogioni, Chieh-Hsin Lai, Yuhta Takida, and Yuki Mitsufuji. VCT: Training consistency models with variational noise coupling. InForty-second International Conference on Machine Learning, 2025. URLhttps://openreview.net/ forum?id=CMoX0BEsDs

    work page 2025

  44. [45]

    Multisample flow matching: Straightening flows with minibatch couplings

    Aram-Alexandre Pooladian, Heli Ben-Hamu, Carles Domingo-Enrich, Brandon Amos, Yaron Lipman, and Ricky TQ Chen. Multisample flow matching: Straightening flows with minibatch couplings. InInternational Conference on Machine Learning, pages 28100–28127. PMLR, 2023

    work page 2023

  45. [46]

    Hierarchical rectified flow matching with mini-batch couplings.arXiv preprint arXiv:2507.13350, 2025

    Yichi Zhang, Yici Yan, Alex Schwing, and Zhizhen Zhao. Hierarchical rectified flow matching with mini-batch couplings.arXiv preprint arXiv:2507.13350, 2025

    work page arXiv 2025

  46. [47]

    Stochastic gradient descent tricks

    Léon Bottou. Stochastic gradient descent tricks. InNeural networks: tricks of the trade: second edition, pages 421–436. Springer, 2012

    work page 2012

  47. [48]

    Elucidating the design space of diffusion-based generative models.Advances in neural information processing systems, 35:26565–26577, 2022

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models.Advances in neural information processing systems, 35:26565–26577, 2022. 12

    work page 2022

  48. [49]

    Learning representations by back-propagating errors.nature, 323(6088):533–536, 1986

    David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. Learning representations by back-propagating errors.nature, 323(6088):533–536, 1986

    work page 1986

  49. [50]

    Scaling Rectified Flow Transformers for High-Resolution Image Synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, Kyle Lacey, Alex Goodwin, Yannik Marek, and Robin Rombach. Scaling rectified flow transformers for high-resolution image synthesis, 2024. URLhttps://arxiv.org/abs/2403.03206

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  50. [51]

    Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers

    Nanye Ma, Mark Goldstein, Michael S Albergo, Nicholas M Boffi, Eric Vanden-Eijnden, and Saining Xie. Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers. InEuropean Conference on Computer Vision, pages 23–40. Springer, 2024

    work page 2024

  51. [52]

    Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think

    Sihyun Yu, Sangkyung Kwak, Huiwon Jang, Jongheon Jeong, Jonathan Huang, Jinwoo Shin, and Saining Xie. Representation alignment for generation: Training diffusion transformers is easier than you think.arXiv preprint arXiv:2410.06940, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  52. [53]

    Reconstruction vs

    Jingfeng Yao, Bin Yang, and Xinggang Wang. Reconstruction vs. generation: Taming optimization dilemma in latent diffusion models. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 15703–15712, 2025

    work page 2025

  53. [54]

    Diffusion Transformers with Representation Autoencoders

    Boyang Zheng, Nanye Ma, Shengbang Tong, and Saining Xie. Diffusion transformers with representation autoencoders.arXiv preprint arXiv:2510.11690, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  54. [55]

    Adversarial Flow Models

    Shanchuan Lin, Ceyuan Yang, Zhijie Lin, Hao Chen, and Haoqi Fan. Adversarial flow models.arXiv preprint arXiv:2511.22475, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  55. [56]

    Inductive moment matching

    Linqi Zhou, Stefano Ermon, and Jiaming Song. Inductive moment matching. In Forty-second International Conference on Machine Learning, 2025. URLhttps:// openreview.net/forum?id=pwNSUo7yUb

    work page 2025

  56. [57]

    UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    Leland McInnes, John Healy, and James Melville. Umap: Uniform manifold approx- imation and projection for dimension reduction.arXiv preprint arXiv:1802.03426, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  57. [58]

    Gradient-based learning applied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 2002

    Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. Gradient-based learning applied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 2002

    work page 2002

  58. [59]

    Auto-Encoding Variational Bayes

    Diederik P Kingma and Max Welling. Auto-encoding variational bayes.arXiv preprint arXiv:1312.6114, 2013

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  59. [60]

    Karl Pearson. Liii. on lines and planes of closest fit to systems of points in space.The London, Edinburgh, and Dublin philosophical magazine and journal of science, 2(11): 559–572, 1901

    work page 1901

  60. [61]

    Analysis of a complex of statistical variables into principal components

    Harold Hotelling. Analysis of a complex of statistical variables into principal components. Journal of educational psychology, 24(6):417, 1933

    work page 1933

  61. [62]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4401–4410, 2019

    work page 2019

  62. [63]

    Adversarial latent autoencoders

    Stanislav Pidhorskyi, Donald A Adjeroh, and Gianfranco Doretto. Adversarial latent autoencoders. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14104–14113, 2020

    work page 2020

  63. [64]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium.Advances in neural information processing systems, 30, 2017

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium.Advances in neural information processing systems, 30, 2017

    work page 2017

  64. [65]

    Imagenet classification with deep convolutional neural networks.Advances in neural information processing systems, 25, 2012

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks.Advances in neural information processing systems, 25, 2012. 13

    work page 2012

  65. [66]

    High-Resolution Image Synthesis with Latent Diffusion Models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models, 2022. URLhttps:// arxiv.org/abs/2112.10752

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  66. [67]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16000–16009, 2022

    work page 2022

  67. [68]

    Improved techniques for training gans.Advances in neural information processing systems, 29, 2016

    Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, and Xi Chen. Improved techniques for training gans.Advances in neural information processing systems, 29, 2016

    work page 2016

  68. [69]

    What Matters in Learning from Offline Human Demonstrations for Robot Manipulation

    Ajay Mandlekar, Danfei Xu, Josiah Wong, Soroush Nasiriany, Chen Wang, Rohun Kulkarni, Li Fei-Fei, Silvio Savarese, Yuke Zhu, and Roberto Martín-Martín. What matters in learning from offline human demonstrations for robot manipulation.arXiv preprint arXiv:2108.03298, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  69. [70]

    Implicit behavioral cloning

    Pete Florence, Corey Lynch, Andy Zeng, Oscar A Ramirez, Ayzaan Wahid, Laura Downs, Adrian Wong, Johnny Lee, Igor Mordatch, and Jonathan Tompson. Implicit behavioral cloning. InConference on robot learning, pages 158–168. PMLR, 2022

    work page 2022

  70. [71]

    Behavior transformers: Cloningk modes with one stone.Advances in neural information processing systems, 35:22955–22968, 2022

    Nur Muhammad Shafiullah, Zichen Cui, Ariuntuya Arty Altanzaya, and Lerrel Pinto. Behavior transformers: Cloningk modes with one stone.Advances in neural information processing systems, 35:22955–22968, 2022

    work page 2022

  71. [72]

    Relay policy learning: Solving long-horizon tasks via imitation and reinforcement learning.arXiv preprint arXiv:1910.11956, 2019

    Abhishek Gupta, Vikash Kumar, Corey Lynch, Sergey Levine, and Karol Hausman. Relay policy learning: Solving long-horizon tasks via imitation and reinforcement learning.arXiv preprint arXiv:1910.11956, 2019

    work page arXiv 1910

  72. [73]

    Diffusion policy: Visuomotor policy learning via action diffusion.The International Journal of Robotics Research, 44(10-11):1684–1704, 2025

    Cheng Chi, Zhenjia Xu, Siyuan Feng, Eric Cousineau, Yilun Du, Benjamin Burchfiel, Russ Tedrake, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion.The International Journal of Robotics Research, 44(10-11):1684–1704, 2025

    work page 2025

  73. [74]

    Stochastic interpolants via conditional dependent coupling.arXiv preprint arXiv:2509.23122, 2025

    Chenrui Ma, Xi Xiao, Tianyang Wang, Xiao Wang, and Yanning Shen. Stochastic interpolants via conditional dependent coupling.arXiv preprint arXiv:2509.23122, 2025

    work page arXiv 2025

  74. [75]

    CAD-VAE: Leveraging correlation-aware latents for comprehensive fair disentanglement

    Chenrui Ma, Xi Xiao, Tianyang Wang, Xiao Wang, and Yanning Shen. CAD-VAE: Leveraging correlation-aware latents for comprehensive fair disentanglement. InThe Fortieth AAAI Conference on Artificial Intelligence, 2025

    work page 2025

  75. [76]

    Beyond editing pairs: Fine- grained instructional image editing via multi-scale learnable regions.arXiv preprint arXiv:2505.19352, 2025

    Chenrui Ma, Xi Xiao, Tianyang Wang, and Yanning Shen. Beyond editing pairs: Fine- grained instructional image editing via multi-scale learnable regions.arXiv preprint arXiv:2505.19352, 2025. 14 Appendix Contents A Preliminary 15 A.1 Flow Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 A.2 Drift Method . . . . . . . . . . ...

    work page arXiv 2025