Synthesizing scientific literature with retrieval-augmented language models

A specialized, open-source, retrieval-augmented language model is introduced for answering scientific queries and synthesizing literature, the responses of which are shown to be preferred by human evaluations over expert-written answers.