# Continuous Adversarial MeanFlow for Few-Step Generative Adaptation from Any Parameterization

DevFeed: [Continuous Adversarial MeanFlow for Few-Step Generative Adaptation from Any Parameterization](<https://devfeed.tech/articles/continuous-adversarial-meanflow-for-few-step-generative-adaptation-from-any-parameterization-120474.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2608.19540>)

Author: Yara Bahram, Zahra Dehghani, M\\'elodie Desbos, Eric Granger, Pablo Piantanida, Mohammadhadi Shateri

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Generative Adversarial Network](<https://devfeed.tech/topics/generative-adversarial-network.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [ai-image-generator](<https://devfeed.tech/topics/ai-image-generator.md>)

Tags: [about](<https://devfeed.tech/tags/about.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [acceleration](<https://devfeed.tech/tags/acceleration.md>), [adapting](<https://devfeed.tech/tags/adapting.md>), [adversarial](<https://devfeed.tech/tags/adversarial.md>), [also](<https://devfeed.tech/tags/also.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [based](<https://devfeed.tech/tags/based.md>), [change](<https://devfeed.tech/tags/change.md>), [compare](<https://devfeed.tech/tags/compare.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data](<https://devfeed.tech/tags/data.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [domain](<https://devfeed.tech/tags/domain.md>), [domain-model](<https://devfeed.tech/tags/domain-model.md>), [domains](<https://devfeed.tech/tags/domains.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [existing](<https://devfeed.tech/tags/existing.md>), [fast](<https://devfeed.tech/tags/fast.md>), [generative](<https://devfeed.tech/tags/generative.md>)

## AI overview

MeanFlow-Transfer (MF-T) adapts pretrained diffusion or flow models across parameterizations through a common velocity map while compressing them into few-step MeanFlow generators. Continuous Adversarial MeanFlow (CAMF) scores changes in a learned potential over each predicted interval jump, addressing the limits of endpoint criteria on small jumps and instantaneous-velocity criteria on large jumps. It requires no gradient penalty or Jacobian-vector product, is reported to remain well-conditioned across interval lengths, and recovers the instantaneous criterion as intervals shrink. Across four ImageNet-based source models and five target domains, MF-T + CAMF approached or exceeded fine-tuned many-step teachers on FID and FDD with up to 125 times fewer Neural Function Evaluations; CAMF also improved MF-T few-step FID by about 30% on average and outperformed prior adversarial post-training on iMF at every evaluated NFE.

## Source excerpt

arXiv:2608.19540v4 Announce Type: replace Abstract: Training fast generators on new domains with limited data remains challenging for two reasons. First, adapting a pretrained diffusion or flow model leaves its costly multi-step sampling intact, and existing acceleration methods are tied to the source parameterization (x, epsilon, v, or u), so heterogeneous pretrained models share no common acceleration target. Second, while adversarial post-training sharpens generation quality, existing criteria each cover only one end of the step range. Discrete criteria compare jump endpoints and lose their signal as the jump shrinks, while continuous criteria score the instantaneous velocity and misjudge the large jumps of few-step sampling. Neither fits MeanFlow (MF) models, which predict average velocities over intervals of any length. We propose MeanFlow-Transfer (MF-T), which adapts a source domain model of any parameterization and compresses it into a few-step MF generator through a common velocity map. We then introduce Continuous Adversarial MeanFlow (CAMF), which scores the change of a learned potential along each predicted interval jump of MF. CAMF needs no gradient penalty or Jacobian-vector product (JVP). We also prove it stays well-conditioned for every interval length and recovers the instantaneous criterion as the interval vanishes. Adapting four ImageNet-based source models (DiT, SiT, JiT, iMF) to five target domains, MF-T + CAMF approaches or exceeds the fine-tuned many-step teacher in FID and FDD at up to 125x fewer Neural Function Evaluations (NFEs). Further, CAMF improves MF-T's few-step FID by about 30% on average, outperforming prior adversarial post-training on iMF at every NFE.