Using Small Models to Compare Language Learning and Tokenizer Performance arxiv.org 1 points by PaulHoule 15 hours ago