AI RESEARCH
Do Heavy Tails Help Diffusion? On the Subtle Trade-off Between Initialization and Training
arXiv CS.LG
•
ArXi:2605.13175v1 Announce Type: new Recent works have proposed incorporating heavy-tailed (HT) noise into diffusion- and flow-based generative models, with the goals of better recovering the tails of target distributions and improving generative diversity. This motivation is intuitive: if the data are heavy-tailed, HT noise may appear better matched than light-tailed (LT) Gaussian noise. However, replacing Gaussian noise by HT noise also changes the underlying estimation problem.