Machine learning potentials and foundation models are changing atomistic simulation — but to truly raise the bar, we need to break out of the standard DFT reference.
Our new paper, published as J. Chem. Theory Comput. (2025) 21, 22, 11710–11720, pushes that frontier: we demonstrate a Δ-learning route towards coupled-cluster accuracy for the condensed phase, closing the gap without prohibitive cost.
Congratulations to this team effort, with special shout-outs to Niamh O’Neill and Benjamin Shi!