Lightweight regularizers cut Fokker-Planck residuals in diffusion models nearly as well as expensive full penalties, at far lower compute cost and with no clear loss in sample quality.
Understanding the latent space of diffusion models through the lens of riemannian geometry
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
-
An Analysis of Regularization and Fokker-Planck Residuals in Diffusion Models for Image Generation
Lightweight regularizers cut Fokker-Planck residuals in diffusion models nearly as well as expensive full penalties, at far lower compute cost and with no clear loss in sample quality.