Machine Learning Interatomic Potential– tag –
-
Meta FAIR and Stanford Researchers Successfully Fine-Tune UMA Model for High-Precision Simulation of WS2 Oxygen Plasma Interactions
arXiv USA Overview Researchers from Meta FAIR and Stanford University successfully fine-tuned the UMA universal machine-learned interatomic potential (MLIP) model specifically for oxygen plasma interactions with WS2. This study addresses... -
Tohoku University Pioneers AI-Powered Closed-Loop System for Rapid Clean Energy Material Discovery
Tohoku University Japan Overview Researchers at Tohoku University have developed an innovative 'closed-loop' system integrating AI models, autonomous experimentation, and continuous feedback into a unified research platform. This framewo... -
MIT Researchers Uncover Critical Role of Coverage-Dependent Lateral Interactions in High-Entropy Alloy Electrocatalytic Activity for Oxygen Reduction Reaction via MLIPs
PubMed USA Overview MIT researchers developed a framework utilizing Machine Learning Interatomic Potentials (MLIPs) to model the oxygen reduction reaction (ORR) within the compositional space of Ag-Ir-Ru-Pd-Pt-Cu-Rh-Re high-entropy alloy... -
UNCC Research Group Develops ‘PyXtal_FF’ for ML Interatomic Potential Generation, Dramatically Reducing Computational Cost of Atomic Simulations
University of North Carolina at Charlotte USA Overview Researchers at the University of North Carolina at Charlotte (UNCC) have developed 'PyXtal_FF,' a package for generating Machine Learning Interatomic Potentials (MLIAP). MLIAPs enabl... -
ML Interatomic Potentials Uncover Order-Disorder Transition and Non-Monotonic Stiffness in (MoCrTi)$_{100-x}$Al$_x$ Refractory High-Entropy Alloys
arXiv International Overview This research leverages universal machine learning interatomic potentials combined with hybrid Monte Carlo and molecular dynamics simulations to investigate the chemical ordering and mechanical properties of ... -
Active Learning for MACE-MP-0 Foundation MLFFs Achieves Full Data Accuracy with Significantly Fewer Labeled Training Examples
arXiv International Overview This study demonstrates an innovative active learning strategy for fine-tuning machine-learning force fields (MLFFs), specifically focusing on foundation models like MACE-MP-0, to achieve full-data accuracy w... -
Pauli Charges Dramatically Enhance Accuracy and High-Pressure, High-Temperature Stability of Machine-Learned Interatomic Potentials
ChemRxiv International Overview This research significantly improves machine-learned interatomic potentials (MLIPs) by integrating Pauli repulsion into a combination of machine-learned pair potentials and short-range atomic neural networ... -
UniFFBench Reveals Critical Role of System-Specific Fine-Tuning for Universal ML Force Fields via Rigorous Experimental Benchmarking of 6 EGraFF Algorithms
ResearchGate International Overview The UniFFBench study rigorously evaluates universal machine learning interatomic potentials (uMLIPs) from six prominent EGraFF algorithms—NequIP, Allegro, BOTNet, MACE, Equiformer, and TorchMDNet—again... -
GPUMD 4.0 Achieves High-Performance Versatile Materials Simulations with Integrated Machine-Learned Potentials
Materials Genome Engineering Advances Global Overview The high-performance molecular dynamics package GPUMD 4.0 has been released, integrating advanced machine-learning potentials (MLPs) based on the NeuroEvolution Potential (NEP) framew... -
arXiv: New ‘ATR’ Framework Enhances Dynamic Stability of MLIPs by Actively Rejecting Pseudo-Labels
arXiv Global Overview A new arXiv preprint introduces the 'Adaptive Multi-Teacher Routing (ATR)' framework, which significantly improves the reliability and generalization capability of universal machine-learning interatomic potentials (...
