{"slug": "continuous-adversarial-meanflow-transfer", "title": "Continuous Adversarial MeanFlow Transfer", "summary": "Researchers propose MeanFlow-Transfer (MF-T) and Continuous Adversarial MeanFlow (CAMF) to adapt pretrained diffusion or flow models to new domains with limited data while accelerating sampling. MF-T maps heterogeneous source outputs into a shared velocity representation and initializes a MeanFlow generator, unifying adaptation and acceleration. CAMF extends adversarial refinement to finite-interval average velocities, improving few-step FID by 29% on average and matching or exceeding fine-tuned teachers at up to 125× fewer NFEs across four ImageNet-based source models and five target domains.", "body_md": "arXiv:2608.19540v1 Announce Type: new\nAbstract: Training fast generators on new domains with limited data remains challenging for two reasons. First, adapting a pretrained diffusion or flow model to a new domain leaves its costly multi-step sampling unaddressed, and existing acceleration methods are tied to the source parameterization--$\\epsilon$, $x$, $v$, or $u$--leaving heterogeneous pretrained models with no common acceleration target. Second, while adversarial refinement is proven effective for few-step quality, it is formulated only for instantaneous-velocity flows, not for the finite-interval average velocities that MeanFlow (MF) models predict. We address both problems. We propose MeanFlow-Transfer, which maps heterogeneous source outputs into a shared velocity representation, uses it to initialize an MF generator from the source weights, and optimizes an MF objective on the target domain. This unifies adaptation and acceleration in a single training loop across a broad range of pretrained models. We then introduce Continuous Adversarial MeanFlow, a post-training stage that extends continuous adversarial flow models from instantaneous velocities to MF's finite-interval average velocities. CAMF contrasts changes in a learned potential between real and predicted interval endpoints, recovering fine detail that MF regression averages away, and reduces to the instantaneous criterion in the vanishing-interval limit. Adapting four ImageNet-based source models--DiT ($\\epsilon$), SiT ($v$), JiT ($x$), iMF ($u$)--to five target domains, MF-T with CAMF matches or exceeds the fine-tuned teacher in FID and FDD at up to $125\\times$ fewer Neural Function Evaluations (NFEs), while CAMF improves MF-T's few-step FID by $29\\%$ on average.", "url": "https://wpnews.pro/news/continuous-adversarial-meanflow-transfer", "canonical_source": "https://arxiv.org/abs/2608.19540", "published_at": "2026-08-21 04:00:00+00:00", "updated_at": "2026-08-21 04:16:17.592752+00:00", "lang": "en", "topics": ["machine-learning", "generative-ai", "artificial-intelligence"], "entities": ["MeanFlow-Transfer", "Continuous Adversarial MeanFlow", "DiT", "SiT", "JiT", "iMF", "ImageNet"], "alternates": {"html": "https://wpnews.pro/news/continuous-adversarial-meanflow-transfer", "markdown": "https://wpnews.pro/news/continuous-adversarial-meanflow-transfer.md", "text": "https://wpnews.pro/news/continuous-adversarial-meanflow-transfer.txt", "jsonld": "https://wpnews.pro/news/continuous-adversarial-meanflow-transfer.jsonld"}}