MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generators

• URL: http://arxiv.org/abs/2607.15273v2
• Date: Tue, 21 Jul 2026 11:14:01 GMT
• Title: MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generators
• Abstract summary: MeanFlowNFT is inspired by the MeanFlow identity, which bridges average and instantaneous velocities.<n>Experiments on image and video generation show that MeanFlowNFT consistently improves baselines.<n>It can even surpass multi-step RL-tuned diffusion while using only a few sampling steps.
• Score: 48.85363546552086
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: MeanFlow generators achieve fast few-step sampling by predicting average velocities over time intervals, making them attractive for efficient generation. Reinforcement learning (RL) has become a powerful way to align diffusion and flow models with human preferences and task-specific objectives. In particular, DiffusionNFT offers an efficient forward-process RL framework that does not require reverse-process trajectories or likelihood estimation. However, applying such RL methods to MeanFlow remains underexplored. DiffusionNFT optimizes instantaneous velocities, whereas MeanFlow samples with average velocities. To bridge this gap, we introduce MeanFlowNFT. Inspired by the MeanFlow identity, which bridges average and instantaneous velocities, we construct an induced instantaneous-velocity predictor. We apply the DiffusionNFT objective to this predictor, making reward optimization well-defined for MeanFlow. Sampling remains based on the average velocity, preserving MeanFlow's fast few-step generation. We further prove that MeanFlowNFT inherits DiffusionNFT's strict policy-improvement guarantee. Experiments on image and video generation show that MeanFlowNFT consistently improves baselines. Moreover, it outperforms prior state-of-the-art RL-tuned few-step generators on most metrics ($6$ of $8$ on SD3.5-M), and can even surpass multi-step RL-tuned diffusion while using only a few sampling steps. For instance, on Wan 2.1, $4$-step MeanFlowNFT reaches a VBench score of $84.33$, surpassing $50$-step LongCat-Video RL ($82.57$).

Related papers

• Continuous Adversarial MeanFlow Transfer [30.946652462509068]
  Adversarial refinement is proven effective for few-step quality, but not for the finite-interval average velocities that MeanFlow (MF) models predict.<n>We propose MeanFlow-Transfer, which maps heterogeneous source outputs into a shared velocity representation.<n>We then introduce Continuous Adversarial MeanFlow, a post-training stage that extends adversarial flow models from instantaneous velocities to MF's finite-interval average velocities.
  arXiv  Detail & Related papers  (2026-08-20T01:20:31Z)
• Score-Based One-step MeanFlow Policy Optimization [4.94481688445056]
  MeanFlow offers a promising alternative by learning an average velocity field that maps noise to data in a single network evaluation.<n>We propose Score-Based One-step MeanFlow Policy Optimization (SOM), an actor-critic algorithm that resolves this by constructing the target velocity field directly from the Q-function.<n>SOM achieves state-of-the-art performance on locomotion tasks with a single generation step, while substantially reducing both training and inference time compared to prior diffusion- and flow-matching-based policies.
  arXiv  Detail & Related papers  (2026-05-22T08:28:51Z)
• Riemannian MeanFlow for One-Step Generation on Manifolds [54.09734511705173]
  Flow Matching enables simulation-free training of generative models.<n>MeanFlow can be extended to manifold-valued generation where velocities lie in location-dependent tangent spaces.<n>Experiments on spheres, tori, and SO(3) demonstrate competitive one-step sampling with improved quality-efficiency trade-offs and substantially reduced sampling cost.
  arXiv  Detail & Related papers  (2026-03-11T12:41:46Z)
• RMFlow: Refined Mean Flow by a Noise-Injection Step for Multimodal Generation [12.979642182577157]
  Mean flow (MeanFlow) enables efficient, high-fidelity image generation, yet its single-function evaluation (1-NFE) generation often cannot yield compelling results.<n>We introduce RMFlow, an efficient multimodal generative model that integrates a coarse 1-NFE MeanFlow transport with a tailored noise-injection refinement step.<n> RMFlow achieves near state-of-the-art results on text-to-image, context-to-molecule, and time-series generation using only 1-NFE, at a computational cost comparable to the baseline MeanFlows.
  arXiv  Detail & Related papers  (2026-01-31T18:27:05Z)
• Flow Straighter and Faster: Efficient One-Step Generative Modeling via MeanFlow on Rectified Trajectories [14.36205662558203]
  Rectified MeanFlow is a framework that models the mean velocity field along the rectified trajectory using only a single reflow step.<n>Experiments on ImageNet at 64, 256, and 512 resolutions show that Re-MeanFlow consistently outperforms prior one-step flow distillation and Rectified Flow methods in both sample quality and training efficiency.
  arXiv  Detail & Related papers  (2025-11-28T16:50:08Z)
• Understanding, Accelerating, and Improving MeanFlow Training [64.84964628592418]
  MeanFlow promises high-quality generative modeling in few steps, by jointly learning instantaneous and average velocity fields.<n>We analyze the interaction between the two velocities and find: (i) well-established instantaneous velocity is a prerequisite for learning average velocity.<n>We design an effective training scheme that accelerates the formation of instantaneous velocity, then shifts emphasis from short to long-interval average velocity.
  arXiv  Detail & Related papers  (2025-11-24T12:59:27Z)
• AlphaFlow: Understanding and Improving MeanFlow Models [74.64465762009475]
  We show that the MeanFlow objective naturally decomposes into two parts: trajectory flow matching and trajectory consistency.<n>Motivated by these insights, we introduce $alpha$-Flow, a broad family of objectives that unifies trajectory flow matching, Shortcut Model, and MeanFlow.<n>When trained from scratch on class-conditional ImageNet-1K 256x256 with vanilla DiT backbones, $alpha$-Flow consistently outperforms MeanFlow across scales and settings.
  arXiv  Detail & Related papers  (2025-10-23T17:45:06Z)
• MeanFlowSE: one-step generative speech enhancement via conditional mean flow [13.437825847370442]
  MeanFlowSE is a conditional generative model that learns the average velocity over finite intervals along a trajectory.<n>On VoiceBank-DEMAND, the single-step model achieves strong intelligibility, fidelity, and perceptual quality with substantially lower computational cost than multistep baselines.
  arXiv  Detail & Related papers  (2025-09-18T11:24:47Z)
• Towards High-Order Mean Flow Generative Models: Feasibility, Expressivity, and Provably Efficient Criteria [14.37317011636007]
  We introduce a theoretical study on Second-Order MeanFlow, a novel extension that incorporates average acceleration fields into the MeanFlow objective.<n>We first establish the feasibility of our approach by proving that the average acceleration satisfies a generalized consistency condition analogous to first-order MeanFlow.<n>We then characterize its expressivity via circuit complexity analysis, showing that the Second-Order MeanFlow sampling process can be implemented by uniform threshold circuits.
  arXiv  Detail & Related papers  (2025-08-09T21:10:58Z)
• Mean Flows for One-step Generative Modeling [64.4997821467102]
  We propose a principled and effective framework for one-step generative modeling.<n>A well-defined identity between average and instantaneous velocities is derived and used to guide neural network training.<n>Our method, termed the MeanFlow model, is self-contained and requires no pre-training, distillation, or curriculum learning.
  arXiv  Detail & Related papers  (2025-05-19T17:59:42Z)
• PeRFlow: Piecewise Rectified Flow as Universal Plug-and-Play Accelerator [73.80050807279461]
  Piecewise Rectified Flow (PeRFlow) is a flow-based method for accelerating diffusion models. PeRFlow achieves superior performance in a few-step generation.
  arXiv  Detail & Related papers  (2024-05-13T07:10:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.