Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow
March 2026
20 min read
Generative Models, Rectified Flow, Domain Transfer, ODE
Use arrow keys or click to navigate slides. Press 'F' or Fullscreen icon for best experience.
What You'll Learn
- •Balancing sample quality vs inference speed
- •Limitations of Diffusion, CNFs & GANs
- •Rectified flow and path rewiring
- •PF-ODE categories: VP, sub-VP, VE
- •Learning ODEs with direct smooth interpolation
- •Unifying generative modeling and domain transfer
Key Concepts Covered
Learning a mapping from a simple distribution (noise) to complex data (images, motion).
Redirects paths that would intersect by rewiring, preserving both map density and causality.
Decreases transport cost and straightens path which reduces calculation steps and time-error.
Resources
Slide Overview
- The Problem: Modeling High-Dimensional Distributions
- Rectified Flow Mechanism
- PF-ODEs Analysis
- Direct ODE Learning & Interpolation
- Results & Key Advantages