CrystalFormer-CSP: Thinking Fast and Slow for Crystal Structure Prediction

2025/12/20 by Zhendong Cao, Cao, Zhendong, Shigang Ou +3
Chemistry · Computer Science · Materials Science · #Benchmark (surveying) #Computational Drug Discovery Methods #Computational Physics (physics.comp-ph) #Crystal structure prediction #Crystallography and molecular interactions #FOS: Computer and information sciences #FOS: Physical sciences #Field (mathematics) #Generative grammar #Heuristic #Interface (matter) #Machine Learning (cs.LG) #Machine Learning in Materials Science #Materials Science (cond-mat.mtrl-sci) #Reinforcement learning

paper · open access · doi:10.48550/arxiv.2512.18251

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2025/12/20 · openalex created_date 2025/12/24 · openalex updated_date 2026/07/28

Abstract

Crystal structure prediction is a fundamental problem in materials science. We present CrystalFormer-CSP, an efficient framework that unifies data-driven heuristic and physics-driven optimization approaches to predict stable crystal structures for given chemical compositions. The approach combines pretrained generative models for space-group-informed structure generation and a universal machine learning force field for energy minimization. Reinforcement fine-tuning can be employed to further boost the accuracy of the framework. We demonstrate the effectiveness of CrystalFormer-CSP on benchmark problems and showcase its usage via web interface and language model integration.

Citations

• AI-Driven Expansion and Application of the Alexandria Database
• Equivariant Diffusion for Crystal Structure Prediction
• OpenCSP: A Deep Learning Framework for Crystal Structure Prediction from Ambient to High Pressure
• RL's Razor: Why Online Reinforcement Learning Forgets Less
• High-dimensional Optimization with Low Rank Tensor Sampling and Local Search
• Self-Optimizing Machine Learning Potential Assisted Automated Workflow for Highly Efficient Complex Systems Material Design
• Orb-v3: atomistic simulation at scale
• Reinforcement Fine-Tuning for Materials Design
• Discovery of High-Temperature Superconducting Ternary Hydrides via Deep Learning
• CrystalFlow: A Flow-Based Generative Model for Crystalline Materials
• Data-efficient multi-fidelity training for high-fidelity machine\n learning interatomic potentials
• MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures
• Space Group Informed Transformer for Crystalline Materials Generation
• Language Modeling Is Compression
• Crystal Structure Prediction by Joint Equivariant Diffusion
• Shotgun crystal structure prediction using machine-learned formation energies
• PROTES: Probabilistic Optimization with Tensor Sampling
• Classifier-Free Diffusion Guidance
• Training language models to follow instructions with human feedback
• Crystal Diffusion Variational Autoencoder for Periodic Material Generation
• Crystal structure prediction at finite temperatures
• Proximal Policy Optimization Algorithms
• Die Gesetze der Krystallochemie

Related

• CSPBench: a benchmark and critical evaluation of Crystal Structure Prediction
• MatLLMSearch: Crystal Structure Discovery with Evolution-Guided Large Language Models
• CrySPAI: A new Crystal Structure Prediction Software Based on Artificial Intelligence
• Predictive crystallography at scale: mapping, validating, and learning from 1000 crystal energy landscapes
• Towards Quantitative Evaluation of Crystal Structure Prediction Performance