Rectified Flow

Cross-source consensus on Rectified Flow from 1 sources and 5 claims.

1 sources · 5 claims

How it works

On the Checkerboard benchmark, vanilla RectFlow achieves SWD near 0.166 at both k=1 and k=2, no better than the base Flow Matching model, confirming that the bottleneck is the coupling integrator. — Divergence-Suppressing Couplings for Rectified Flow
In practice, the compounding improvement of iterative rectification fails because the bottleneck is the quality of the coupling integrator, not the loss function or model capacity. — Divergence-Suppressing Couplings for Rectified Flow
Image generation experiments evaluate only k=1 reflow because training costs scale with the number of rounds, leaving multi-round rectification on images undemonstrated. — Divergence-Suppressing Couplings for Rectified Flow
Rectified Flow addresses trajectory curvature by iteratively regenerating coupling data through forward integration of the previously trained model, then retraining on those self-generated pairs. — Divergence-Suppressing Couplings for Rectified Flow
Each successive reflow round is expected to produce a velocity field closer to linear, enabling few-step or single-step generation. — Divergence-Suppressing Couplings for Rectified Flow