Mixing unconditional Gaussian noise with a κ-conditioned source during training of rectified flows reduces path curvature, yielding 12% better FID scores and faster sampling than standard rectified flows.
Learning multiple layers of features from tiny images
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
MixFlow: Mixed Source Distributions Improve Rectified Flows
Mixing unconditional Gaussian noise with a κ-conditioned source during training of rectified flows reduces path curvature, yielding 12% better FID scores and faster sampling than standard rectified flows.