What question did this study set out to answer?

The aim is to enhance prediction accuracy for material properties using advanced modeling techniques.

May 22, 2026Open Access

Enhancing composition-based materials property prediction

Key Points

The aim is to enhance prediction accuracy for material properties using advanced modeling techniques.
Proposed implicit transfer through pretraining chemical language models on multimodal embeddings.
Introduced explicit transfer that generates crystal structures with structure-aware predictors.
Benchmarking performed on LLM4Mat-Bench and MatBench tasks.
Achieved state-of-the-art performance in 25 out of 32 benchmark cases.
Demonstrated improved model interpretability using a game-theoretic approach.
Incorporated high-order feature interactions effectively in the prediction models.

Abstract

Abstract Crystal graph neural networks are widely applicable in modeling experimentally synthesized compounds and hypothetical materials with unknown synthesizability. In contrast, structure-agnostic predictive algorithms allow exploring previously inaccessible domains of chemical space. Here we present a universal approach for enhancing composition-based materials property prediction by means of cross-modal knowledge transfer. Two formulations are proposed: implicit transfer involves pretraining chemical language models on multimodal embeddings, whereas explicit transfer suggests generating crystal structures and implementing structure-aware predictors. The proposed approaches were benchmarked on LLM4Mat-Bench and MatBench tasks, achieving state-of-the-art performance in 25 out of 32 cases. In addition, we demonstrated how another modeling aspect of chemical language models—interpretability—benefits from applying a game-theoretic approach, which is able to incorporate high-order feature interactions.

Mark Helpful

Bookmark

Relay

View Full Paper