Gradient Variance Reveals Failure Modes in Flow-Based Generative Models

NeurIPSSpotlight2025

Authors
Teodora Reu, Sixtine Dromigny, Michael M. Bronstein, Francisco Vargas
Venue
NeurIPS 2025
Track
Spotlight

TL;DR

Rectified Flows learn ODE vector fields whose trajectories are straight between source and target distributions, enabling near one-step inference…

Opening excerpt from the authors’ abstract. source

Read the paper

Topics

generative model

Related NeurIPS papers

← All NeurIPS 2025 Spotlight papers · Browse the whole archive