AI and the World's Hard Problems · Chemistry & materials · arXiv · 2026-10-20
Crystal Structure Based Graph Neural Network And LLM Multi-Modal Approach for Material Magnetic Ordering Prediction
The authors propose a multi-modal model to classify a material's magnetic ordering into four classes. A graph neural network takes the crystal structure, and an LLM takes text descriptions of the material, and their outputs are fused. The combined approach achieved higher overall accuracy than either the GNN or the LLM alone across three fusion strategies.