Representations from Pretrained Machine-Learning Interatomic Potentials as Coarse Coordinates for Material Generation and Evaluation

2026-07-30

AI Summary

Existing generative models for inorganic crystals often rely on simplified structural representations, struggling to balance generation quality and novelty. This work proposes a coarse-grained atomic feature representation derived from pretrained machine-learned interatomic potentials—such as MACE—and introduces a novel evaluation metric, the Coarse-to-Fine Transport Distance (CFTD), grounded in optimal transport theory. CFTD provides a unified assessment of both structural quality and novelty while effectively detecting model memorization. Experimental results demonstrate that CFTD outperforms existing metrics, such as SUN, in evaluation fidelity and can be seamlessly integrated into generative frameworks, significantly enhancing the diversity and chemical plausibility of generated crystal structures.

Abstract

Generative machine learning is increasingly used for inorganic crystal structure generation. Most models and the corresponding evaluation approaches rely on simple forms of crystal structure representation. In this paper, we showcase the power of atom-averaged features from pretrained Machine-Learning Interatomic Potentials (MLIPs), such as MACE, for such tasks. We first introduce a distance measure that assesses the output of material generative models by capturing both quality and novelty in a single distribution-based evaluation framework. In particular, we introduce the Coarse-Fine Transport Distance (CFTD) using two different featurizers, where the quality component is based on coarse MACE features. We showcase CFTD's versatility in capturing crystal-structure quality while also detecting memorization, and compare it with the recently introduced continuous SUN metrics. We further show that coarse MACE features can be used as guidance for a material generative model.

Problem

• crystal structure generation
• material representation
• generative models
• evaluation metrics
• novelty and quality assessment

Innovation

• Machine-Learning Interatomic Potentials
• Coarse-Fine Transport Distance
• Material Generation
• Crystal Structure Representation
• Generative Evaluation