Equivariant Asynchronous Diffusion: An Adaptive Denoising Schedule for Accelerated Molecular Conformation Generation

arXiv cs.AI / 3/12/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

Equivariant Asynchronous Diffusion (EAD) introduces a diffusion model that blends asynchronous denoising with a molecule-level horizon to more effectively model hierarchical molecular structures.
It uses a dynamic, adaptive denoising schedule to determine timesteps, aiming to bridge training-inference gaps and capture causal relationships in molecular hierarchies.
By combining strengths of auto-regressive and diffusion approaches, EAD seeks to overcome limitations of existing methods in 3D molecular generation.
Experimental results show state-of-the-art performance on 3D molecular generation benchmarks, highlighting its potential for accelerated molecular design.

Abstract

Recent 3D molecular generation methods primarily use asynchronous auto-regressive or synchronous diffusion models. While auto-regressive models build molecules sequentially, they're limited by a short horizon and a discrepancy between training and inference. Conversely, synchronous diffusion models denoise all atoms at once, offering a molecule-level horizon but failing to capture the causal relationships inherent in hierarchical molecular structures. We introduce Equivariant Asynchronous Diffusion (EAD) to overcome these limitations. EAD is a novel diffusion model that combines the strengths of both approaches: it uses an asynchronous denoising schedule to better capture molecular hierarchy while maintaining a molecule-level horizon. Since these relationships are often complex, we propose a dynamic scheduling mechanism to adaptively determine the denoising timestep. Experimental results show that EAD achieves state-of-the-art performance in 3D molecular generation.