Fine-tuning– tag –
-
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... -
arXiv Paper Demonstrates Full-Data Accuracy with Fewer Labels in ML Force Field Training Through Data Selection Strategy
arXiv International Overview This paper investigates the critical role of data selection in training and fine-tuning machine-learning force fields (MLFFs), demonstrating that active learning strategies like LLPR (Least-Likely to Predict ... -
Foundation Models Uncover Novel High-Pressure Phase Ca6FeNi, Revolutionizing Materials Discovery Workflow
arXiv USA Overview A recent preprint on arXiv introduces a self-consistent foundation model-assisted crystal structure prediction (CSP) workflow that integrates evolutionary search with adaptive data selection and fine-tuning. This innov... -
LLMs Drive Autonomous Research to Boost Material Bandgap Predictions
arXiv USA Overview A recent preprint introduces an autonomous research loop powered by large language models (LLMs) that significantly enhances the optimization of crystal graph networks for electronic bandgap prediction. This self-consi... -
Bias Identified in Universal Machine-Learned Interatomic Potentials; Iterative Fine-Tuning Improves Accuracy
Journal of Chemical Theory and Computation (ACS Publications) USA Overview This study thoroughly investigated intrinsic biases in universal machine-learned interatomic potentials (uMLIPs), such as MACE, and their impact on fine-tuning qu... -
Meta FAIR’s Universal MLIP ‘UMA’ Precisely Models Oxygen Plasma Interactions with 2D Materials, Advancing Semiconductor Manufacturing
arXiv Unknown Overview Meta FAIR's universal machine-learned interatomic potential (MLIP) model, UMA, has demonstrated highly accurate modeling of oxygen plasma interactions with tungsten disulfide (WS2), a 2D material, with performance ... -
arXiv Paper Evaluates Universal MLIPs, Finds DFT Fine-Tuning Essential for Accuracy in Reactive Processes
arXiv USA Overview A new arXiv study evaluated five universal machine-learning interatomic potentials (MLIPs) for quantitative materials modeling involving reactive events. It revealed that while universal MLIPs are becoming general-purp... -
Sparsity-Promoting Fine-Tuning Enhances Domain Adaptability of Pre-Trained Equivariant Materials Foundation Models
arXiv International Overview A sparsity-promoting fine-tuning method has been proposed for robust and interpretable adaptation of pre-trained equivariant materials foundation models (MLIPs) to domain-specific applications. This technique... -
DP-EVA Framework Maximizes Pre-Trained Knowledge of Large Atomistic Models to Develop Data-Efficient MLIPs
Clean Energy | Oxford Academic International Overview A new data-efficient fine-tuning framework, DP-EVA, has been introduced, enabling the development of domain-specific Machine Learning Interatomic Potentials (MLIPs) by maximizing the ...
12