This "was credited with greatly improving our understanding of how large language models (LLMs) work as well as providing a roadmap of how to build the next generation of these technologies." (from ACM profile of author Sewon Min)