Saved in:
Bibliographic Details
Main Authors: Ni, Yuyan, Feng, Shikun, Ma, Wei-Ying, Ma, Zhi-Ming, Lan, Yanyan
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.17340
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911016375287808
author Ni, Yuyan
Feng, Shikun
Ma, Wei-Ying
Ma, Zhi-Ming
Lan, Yanyan
author_facet Ni, Yuyan
Feng, Shikun
Ma, Wei-Ying
Ma, Zhi-Ming
Lan, Yanyan
contents Sampling strategies in diffusion models are critical to molecular generation yet remain relatively underexplored. In this work, we investigate a broad spectrum of sampling methods beyond conventional defaults and reveal that sampling choice substantially affects molecular generation performance. In particular, we identify a maximally stochastic sampling (StoMax), a simple yet underexplored strategy, as consistently outperforming default sampling methods for generative models DDPM and BFN. Our findings highlight the pivotal role of sampling design and suggest promising directions for advancing molecular generation through principled and more expressive sampling approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2506_17340
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Revisiting Sampling Strategies for Molecular Generation
Ni, Yuyan
Feng, Shikun
Ma, Wei-Ying
Ma, Zhi-Ming
Lan, Yanyan
Chemical Physics
Sampling strategies in diffusion models are critical to molecular generation yet remain relatively underexplored. In this work, we investigate a broad spectrum of sampling methods beyond conventional defaults and reveal that sampling choice substantially affects molecular generation performance. In particular, we identify a maximally stochastic sampling (StoMax), a simple yet underexplored strategy, as consistently outperforming default sampling methods for generative models DDPM and BFN. Our findings highlight the pivotal role of sampling design and suggest promising directions for advancing molecular generation through principled and more expressive sampling approaches.
title Revisiting Sampling Strategies for Molecular Generation
topic Chemical Physics
url https://arxiv.org/abs/2506.17340