Generative artificial intelligence (AI) systems can be optimized using TextGrad, a framework that performs optimization by backpropagating large-language-model-generated feedback; TextGrad enables optimization across diverse tasks, including radiotherapy treatment plans and molecule generation.
