The latter work proposed to use existing Large Language Models (LLMs) to
generate “textbook quality” data as a way to enhance the learning process compared to traditional
web data.