AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow

ICLR 2023 · Image generation › Fast diffusion sampling · approach: Rectified flow straightening · 84 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add one reflow step to the rectified flow (ablation)Rectified Flow on CIFAR-10 · CIFAR-10FID: 378 -> 12.21Helped
Add one reflow step, from 2- to 3-Rectified Flow (ablation)Rectified Flow on CIFAR-10 · CIFAR-10FID: 12.21 -> 8.15Helped
Add one reflow step to the rectified flow (ablation)Rectified Flow on CIFAR-10 · CIFAR-10IS: 1.13 -> 8.08Helped
Add one reflow step, from 2- to 3-Rectified Flow (ablation)Rectified Flow on CIFAR-10 · CIFAR-10IS: 8.08 -> 8.47Helped
Use a nonlinear interpolation instead of the canonical linear interpolation (ablation)Rectified Flow · Distribution transportconvex transport costMixed
Add too many reflow steps (ablation)Rectified Flow · Distribution transportflow estimation accuracyHurt
Increase L2 regularization to smooth the neural velocity model (ablation)Neural Rectified Flow on toy examples · Toy examplesflow straightnessHelped
Replace the linear interpolation schedule with an exponential schedule in VP ODE (ablation)VP ODE · Few-step generationlarge-step performanceHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy