SplitMeanFlow: Interval Splitting Consistency in Few-Step Generative Modeling
read the original abstract
Generative models like Flow Matching have achieved state-of-the-art performance but are often hindered by a computationally expensive iterative sampling process. To address this, recent work has focused on few-step or one-step generation by learning the average velocity field, which directly maps noise to data. MeanFlow, a leading method in this area, learns this field by enforcing a differential identity that connects the average and instantaneous velocities. In this work, we argue that this differential formulation is a limiting special case of a more fundamental principle. We return to the first principles of average velocity and leverage the additivity property of definite integrals. This leads us to derive a novel, purely algebraic identity we term Interval Splitting Consistency. This identity establishes a self-referential relationship for the average velocity field across different time intervals without resorting to any differential operators. Based on this principle, we introduce SplitMeanFlow, a new training framework that enforces this algebraic consistency directly as a learning objective. We formally prove that the differential identity at the core of MeanFlow is recovered by taking the limit of our algebraic consistency as the interval split becomes infinitesimal. This establishes SplitMeanFlow as a direct and more general foundation for learning average velocity fields. From a practical standpoint, our algebraic approach is significantly more efficient, as it eliminates the need for JVP computations, resulting in simpler implementation, more stable training, and broader hardware compatibility. One-step and two-step SplitMeanFlow models have been successfully deployed in large-scale speech synthesis products (such as Doubao), achieving speedups of 20x.
This paper has not been read by Pith yet.
Forward citations
Cited by 10 Pith papers
-
AnyFlow: Any-Step Video Diffusion Model with On-Policy Flow Map Distillation
AnyFlow enables any-step video diffusion by distilling flow-map transitions over arbitrary time intervals with on-policy backward simulation.
-
Why Are DMD Students Lazy? Understanding the Copying Behavior in Few-Step Distillation
In high-dimensional DMD distillation, student models spontaneously copy teacher noise-data pairings as an emergent effect of limited geometric freedom rather than adversarial objectives or memorization.
-
CoFlow: Coordinated Few-Step Flow for Offline Multi-Agent Decision Making
CoFlow achieves state-of-the-art coordination quality in offline MARL using only 1-3 denoising steps by natively coupling velocity fields across agents via coordinated attention and gating.
-
CoFlow: Coordinated Few-Step Flow for Offline Multi-Agent Decision Making
CoFlow achieves state-of-the-art coordination in offline MARL using single-pass joint velocity fields with Coordinated Velocity Attention and Adaptive Coordination Gating.
-
Noise-Started One-Step Real-World Super-Resolution via LR-Conditioned SplitMeanFlow and GAN Refinement
SMFSR achieves state-of-the-art perceptual quality among one-step diffusion-based real-world super-resolution methods by preserving noise-started generation via LR-conditioned SplitMeanFlow and GAN refinement.
-
CoFlow: Coordinated Few-Step Flow for Offline Multi-Agent Decision Making
CoFlow preserves inter-agent coordination in few-step offline MARL by using a natively joint velocity field with Coordinated Velocity Attention and Adaptive Coordination Gating, matching or exceeding baselines in 1-3 ...
-
Extending One-Step Image Generation from Class Labels to Text via Discriminative Text Representation
By requiring and using highly discriminative LLM text features, the work enables the first effective one-step text-conditioned image generation with MeanFlow.
-
Dual-End Consistency Model
DE-CM reaches state-of-the-art one-step FID of 1.70 on ImageNet 256x256 by decomposing PF-ODE trajectories into three critical sub-trajectories and using flow matching plus N2N mapping for stability.
-
Exposure Bias Can Alleviate Itself via Directional and Frequency Rectification in Flow Matching
DEFAR rectifies exposure bias in Flow Matching by treating bias signals as adaptive feedback for directional correction and low-frequency compensation, outperforming baselines on CIFAR-10, CelebA-64, and ImageNet.
-
CaloTrilogy: Toward a Breakthrough in One-Step, End-to-End, Physics-Guided Shower Generation for Modern Calorimeters
Presents CaloTrilogy, a unified one-step generative model for high-granularity calorimeter showers that combines velocity field integration, learned priors, and physics losses to match SOTA quality.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.