“…We now shift our focus to recent advancements in modelling or the rapid generation of MLIPs for arbitrary chemical systems, with a particular emphasis on disordered melts. In recent work, Guo et al [59] introduced a high-performance active learning workflow termed AL4GAP. This workflow is designed to generate compositionally transferable MLIPs over charge-neutral mixtures of arbitrary molten mixtures, spanning 11 cations (Li, Na, K, Rb, Cs, Mg, Ca, Sr, Ba, Nd, and Th) and 4 anions (F, Cl, Br, I).…”