logoalt Hacker News

a2ff6eeb0yesterday at 9:29 PM0 repliesview on HN

What evals did you use to compare that to using an existing language with a large corpus of training data?