{"id":"arxiv/2610.40308","url":"https://arxiv.org/abs/2610.40308","tags":["math.ST"],"score":0,"title":"Enhanced convergence guarantees of score-based generative models in $\\mathcal{W}_2$-distance beyond log-concavity","domain":"math.ST","source":"arxiv","authors":["Xiaoyan Zhang","Chenxu Pang","Xiaojie Wang"],"reviews":0,"subject":"math.ST","summary":"Although score-based generative models (SGMs) have achieved remarkable success in real-world sample generation tasks, it is still far from sufficient to understand them from a mathematical perspective. In this paper, we aim to provide enhanced convergence guarantees for two existing SGMs in $\\mathcal{W}_2$-distance beyond log-concavity. More precisely, we develop a novel framework of {\\color{black}discretization} error analysis for Euler-Maruyama (EM) and Poisson midpoint (PM) time discretization schemes for SGMs. Under a non-log-concavity condition, we show that…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-29","reactions":{}}