BezierFlow: Learning Bezier Stochastic Interpolant Schedulers for Few-Step Generation

arXiv:2512.13255v3 Announce Type: replace Abstract: We introduce BezierFlow, a lightweight training approach for few-step generation with pretrained diffusion and flow models. BezierFlow achieves a 2-3x performance improvement for sampling with leq 10 NFEs while requiring only 15 minutes of training. Recent lightweight training approaches have shown promise by learning optimal timesteps, but their scope remains restricted to ODE discretizations. To broaden this scope, we propose learning the optimal transformation of the sampling trajectory by parameterizing stochastic interpolant (SI) schedulers. The main challenge lies in designing a parameterization that satisfies critical desiderata, including boundary conditions, differentiability, and monotonicity of the SNR. To effectively meet these requirements, we represent scheduler functions as Bezier functions, where control points naturally enforce these properties. This reduces the problem to learning an ordered set of points in the time range, while the interpretation of the points changes from ODE timesteps to Bezier control points. Across a range of pretrained diffusion and flow models, BezierFlow consistently outperforms prior timestep-learning methods, demonstrating the effectiveness of expanding the search space from discrete timesteps to Bezier-based trajectory transformations.

Related

• Adjoint Matching through the Lens of the Stochastic Maximum Principle in Optimal Control
• Discrete Meanflow Training Curriculum
• Learning Discrete Diffusion of Graphs via Free-Energy Gradient Flows
• TopoFlow: Topography-aware Pollutant Flow Learning for High-Resolution Air Quality Prediction

Source: arXiv cs.LG | 2026-04-14