Guiding Generative Models to Uncover Diverse and Novel Crystals via Reinforcement Learning

2025/11/10 by Park, Hyunsoo, Walsh, Aron · 3 citations
#Computational Physics (physics.comp-ph) #FOS: Computer and information sciences #FOS: Physical sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2511.07158

Abstract

Discovering functional crystalline materials entails navigating an immense combinatorial design space. While recent advances in generative artificial intelligence have enabled the sampling of chemically plausible compositions and structures, a fundamental challenge remains: the objective misalignment between likelihood-based sampling in generative modelling and targeted focus on underexplored regions where novel compounds reside. Here, we introduce a reinforcement learning framework that guides latent denoising diffusion models toward diverse and novel, yet thermodynamically viable crystalline compounds. Our approach integrates group relative policy optimisation with verifiable, multi-objective rewards that jointly balance creativity, stability, and diversity. Beyond de novo generation, we demonstrate enhanced property-guided design that preserves chemical validity, while targeting desired functional properties. This approach establishes a modular foundation for controllable AI-driven inverse design that addresses the novelty-validity trade-off across scientific discovery applications of generative models.

Citations

• Accelerating inverse materials design using generative diffusion models with reinforcement learning
• CrystalGym: A New Benchmark for Materials Discovery Using Reinforcement Learning
• Reinforcement Learning from Human Feedback
• Reinforcement Fine-Tuning for Materials Design
• All-atom Diffusion Transformers: Unified generative modelling of molecules and materials
• A generative material transformer using Wyckoff representation
• CrystalFlow: A Flow-Based Generative Model for Crystalline Materials
• Fine-Tuned Language Models Generate Stable Inorganic Materials as Text
• BioXP-0.5B: Explainable Medical-AI via RL-GRPO
• Large-scale Reinforcement Learning for Diffusion Models
• A foundation model for atomistic materials chemistry
• WyCryst: Wyckoff Inorganic Crystal Generator Framework
• Crystal Structure Prediction by Joint Equivariant Diffusion
• Crystal Structure Generation with Autoregressive Large Language Modeling
• Scalable Diffusion Models with Transformers
• Large-scale machine-learning-assisted exploration of the whole materials space
• Classifier-Free Diffusion Guidance
• High-Resolution Image Synthesis with Latent Diffusion Models
• Crystal Diffusion Variational Autoencoder for Periodic Material Generation
• Structured Denoising Diffusion Models in Discrete State-Spaces
• Denoising Diffusion Implicit Models
• Denoising Diffusion Probabilistic Models
• Improved Techniques for Training Score-Based Generative Models
• Generative Adversarial Networks for Crystal Structure Prediction
• Proximal Policy Optimization Algorithms
• Attention Is All You Need
• A Kernel Method for the Two-Sample Problem
• Efficient iterative schemes forab initiototal-energy calculations using a plane-wave basis set
• Projector augmented-wave method

Cited by

• Generative models for crystalline materials
• Discovery and recovery of crystalline materials with property-conditioned transformers
• Introducing physics-informed generative models for targeting structural novelty in the exploration of chemical space
• Representations from Pretrained Machine-Learning Interatomic Potentials as Coarse Coordinates for Material Generation and Evaluation

Related

• Introducing physics-informed generative models for targeting structural novelty in the exploration of chemical space
• Establishing baselines for generative discovery of inorganic crystals
• Chemical filters for ultra-high-throughput materials screening and generation
• Scalable Diffusion for Materials Generation
• Generative Inverse Design of Crystal Structures via Diffusion Models with Transformers