Derives closed-form optimal coefficient for conditional velocity as control variate in MeanFlow loss, unifying remedies and revealing mismatch between gradient-MSE and FID optima.
Functional mean flow in hilbert space
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
On Variance Reduction in Learning Mean Flows
Derives closed-form optimal coefficient for conditional velocity as control variate in MeanFlow loss, unifying remedies and revealing mismatch between gradient-MSE and FID optima.