AlphaFlow: Understanding and Improving MeanFlow Models
This work improves generative modeling efficiency for researchers and practitioners by addressing a specific bottleneck in MeanFlow, though it is incremental as it builds on existing frameworks.
The paper tackled the optimization conflict in MeanFlow models for few-step generative modeling by introducing α-Flow, a unified objective that disentangles conflicting terms through a curriculum strategy, achieving state-of-the-art FID scores of 2.58 (1-NFE) and 2.15 (2-NFE) on ImageNet-1K 256x256.
MeanFlow has recently emerged as a powerful framework for few-step generative modeling trained from scratch, but its success is not yet fully understood. In this work, we show that the MeanFlow objective naturally decomposes into two parts: trajectory flow matching and trajectory consistency. Through gradient analysis, we find that these terms are strongly negatively correlated, causing optimization conflict and slow convergence. Motivated by these insights, we introduce $α$-Flow, a broad family of objectives that unifies trajectory flow matching, Shortcut Model, and MeanFlow under one formulation. By adopting a curriculum strategy that smoothly anneals from trajectory flow matching to MeanFlow, $α$-Flow disentangles the conflicting objectives, and achieves better convergence. When trained from scratch on class-conditional ImageNet-1K 256x256 with vanilla DiT backbones, $α$-Flow consistently outperforms MeanFlow across scales and settings. Our largest $α$-Flow-XL/2+ model achieves new state-of-the-art results using vanilla DiT backbones, with FID scores of 2.58 (1-NFE) and 2.15 (2-NFE).