MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generators
2026-07-16 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors present MeanFlowNFT, a new method that improves fast data generation models called MeanFlow by combining them with reinforcement learning techniques from DiffusionNFT. They solve a key challenge by connecting average velocities used in MeanFlow with the instantaneous velocities optimized in DiffusionNFT. This allows them to fine-tune the model to produce better images and videos quickly, using only a few steps. Their experiments show that MeanFlowNFT works better than previous few-step generators and even beats some models that use many more steps.
MeanFlowreinforcement learningDiffusionNFTvelocity predictionimage generationvideo generationsampling stepspolicy improvementfew-step generation
Authors
Yushi Huang, Xiangxin Zhou, Jun Zhang, Liefeng Bo, Tianyu Pang
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 ($82.57$).