Large language models as uncertainty-calibrated optimizers for experimental discovery

Bojana Ranković