Stop Typing One Token at a Time: The Coarse-to-Fine Model That's 31 Faster

Every language model you have ever used writes the same way: one token, then the next, then the next,...