Data and software for "Fidelity-Aware Machine-Learned Activation Barriers without Saddle Searches"

Data and software underlying the study “Fidelity-Aware Machine-Learned Activation Barriers without Saddle Searches.” The deposit includes NEB barrier and minimum-energy-path datasets, descriptor vectors and data splits, trained machine-learning models, training scripts, descriptor and path-correction code, KMC simulation data and inputs, and analysis scripts used to produce the reported results. The POD interatomic potential is available separately from its original release.