New AI Method Enables Reinforcement Learning for Crystal Structure Prediction
Researchers introduced OMatG-IRL, a reinforcement learning framework that improves generative models for predicting stable crystal structures by operating directly on velocity fields without requiring explicit score computation. The method addresses a key limitation in applying policy-gradient RL to flow-based generative models used in materials science. This advancement could accelerate materials discovery by enabling faster, more efficient prediction of crystal structures with desired properties.
A new machine learning approach called Open Materials Generation with Inference-time Reinforcement Learning (OMatG-IRL) combines generative models with reinforcement learning to predict crystal structures more effectively. The method works by applying policy-gradient RL directly to velocity fields learned by flow-based models, eliminating the need to compute explicit scores—a technical barrier that previously prevented RL application to this class of models. The framework uses stochastic perturbations to enable exploration and policy learning at inference time while maintaining the baseline performance of pretrained models. In testing, OMatG-IRL achieved results competitive with score-based RL approaches while enabling composition conditioning to preserve diversity. The researchers also demonstrated that the method can learn time-dependent velocity-annealing schedules, achieving order-of-magnitude improvements in sampling efficiency and generation time. The code has been released as part of the Open Materials Generation framework.
What's missing
The paper does not discuss potential limitations of the approach, such as computational requirements for different system sizes, scalability to more complex crystal structures, or validation against experimental data. The study's own scope and generalizability boundaries are not detailed in the abstract.
What different sources said
- arXiv cs.LGCenter
Open Materials Generation with Inference-Time Reinforcement Learning
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.