Crys-JEPA: Accelerating Crystal Discovery via Embedding Screening and Generative Refinement

TL;DR

Crys-JEPA introduces a novel embedding-based approach for crystal generation that balances stability and novelty, significantly improving discovery metrics by reducing reliance on energy evaluations.

Contribution

The paper presents Crys-JEPA, a joint embedding predictive model that captures formation-energy differences, enabling efficient screening and refinement of crystal candidates.

Findings

Achieved up to 81.4% and 82.6% improvements on V.S.U.N metric.

Identified the stability-novelty trade-off in existing models.

Reduced energy evaluation costs through embedding-based stability assessment.

Abstract

De novo crystal generation seeks to discover materials that are not merely realistic, but also stable and novel. However, most existing generative models are trained to maximize the likelihood of observed crystals, which encourages samples to stay close to known materials yet not necessarily align with the criteria that matter in discovery. Through an empirical investigation, we show that current crystal generative models are caught in a pronounced stability--novelty trade-off: moving toward the observed distribution preserves stability but limits novelty, whereas moving away from it quickly destroys stability. This suggests that the useful region for discovering crystals that are both stable and novel is extremely narrow. To escape the trade-off, we introduce Crys-JEPA, a joint embedding predictive architecture for crystals that learns an energy-aware latent space preserving formation-energy differences. In this space, stability assessment can be reformulated as an embedding-based comparison against accessible training crystals, reducing the reliance on expensive energy evaluation and task-specific external references. Building on Crys-JEPA, we further develop a screening-and-refinement pipeline that identifies promising generated crystals and reintroduces them to refine the generative model. On MP-20 and Alex-MP-20 datasets, we achieve improvements over baselines up to 81.4% and 82.6% on V.S.U.N metric, respectively.