"Augmented zero-shot learning" is an awfully clever learning prompt technique for using language models to transform text: give the model a few examples of rewriting text for *other* transforms, then make your request. Surprisingly general performance. arxiv.org/abs/2109.03910
Conversation
Replying to
Hmm, application that jumps to mind is rewriting other ideas from an author or source in your own words - take a few highlights from an article and put them in your own words, then try the style transfer on other passages from the piece/author