DFT-computed Adiabatic Redox Table (DART) Benchmark Study

This repository contains the models, molecular representations, predictions, training data, and analysis code used in the DART benchmarking study. The study compares molecular representations and machine-learning models for predicting neutral HOMO and LUMO energies, adiabatic ionization potentials, and adiabatic electron affinities. The adiabatic ionization potentials (aIP) and adiabatic electron affinities (aEA) are reported in eV; HOMO and LUMO energies are reported in Hartree.