Notes on language models, training data, and evaluation — with room for the things around them.https://kantahayashiai.github.io/