Bochum, Germany, October 29, 2025, Researchers from Research Center for Future Energy Materials and Systems at the Ruhr University Bochum, Software for Chemistry & Materials BV, and Vrije Universiteit Amsterdam have demonstrated that modern universal machine learning interatomic potentials (uMLIPs) can now accurately describe systems ranging from single molecules to bulk solids, representing a significant leap forward for uMLIPs in materials science. The study introduces the 0123D dataset, comprising 40,000 diverse structures specifically designed to benchmark model performance across all dimensionalities.

Researchers from the Technical University of Munich have developed URNet, a novel artificial intelligence model that helps autonomous driving systems perceive their surroundings more clearly—even in dark, fast-changing environments. By combining an unconventional “event camera” with a self-aware framework, URNet allows vehicles to build reliable 3D maps that measure how far objects are—a process known as depth estimation—while understanding how confident they should be about what they “see.” This innovation could make next-generation self-driving cars safer and more capable of navigating complex real-world conditions.

In a paper published in SCIENCE CHINA Earth Sciences, a team of researchers investigated a fine-scale lightning forecasting approach based on weather foundation models (WFMs) and proposed a dual-source data-driven forecasting framework that integrates the strengths of both WFMs and recent lightning observations to enhance predictive performance. Furthermore, a gated spatiotemporal fusion network (gSTFNet) is designed to address the challenges of cross-temporal and cross-modal fusion inherent in dual-source data integration. Experimental results demonstrate that the dual-source framework significantly improves forecasting performance compared to models trained solely on WFMs and outperforms both the ECMWF HRES lightning product and other deep-learning spatiotemporal forecasting models.