Mixture of Experts framework speeds up atomistic simulations

Researchers have developed a new Mixture-of-Experts (MoE) framework for Machine Learning Interatomic Potentials (MLIPs) to accelerate atomistic simulations. This approach divides simulation domains into regions of varying chemical complexity, assigning different model capacities to each. A co-training strategy ensures consistency between models at domain interfaces, preventing artificial stress fields. The framework was validated on a Platinum-Carbon Monoxide system, showing it can double computational speed while maintaining predictive accuracy and energy conservation. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel MoE framework to significantly speed up atomistic simulations for materials science.

RANK_REASON This is a research paper introducing a novel framework for accelerating simulations.

Read on arXiv cs.LG →

• Mixture-of-Experts
• Machine Learning Interatomic Potentials
• MLIPs
• E(3)-equivariant Allegro
• Pt+CO
• arXiv
• Gabriel De Miranda Nascimento

• paper
• infra