Adaptive Transition State Refinement with Learned Equilibrium Flows

Pretrained model and low-fidelity datasets, including: The Transition1x dataset in ase.db format used for training. The OA-Reactdiff data split applied in this work. Files with the suffix _own refer to the same OA-Reactdiff split, but with 1,000 datapoints removed from trainingand used for validation during hyperparameter tuning. Low-fidelity samples as .extxyz database.