1 paper
Buu Phan, Marton Havasi, Matthew Muckley +1
State-of-the-art language models are autoregressive and operate on subword units known as tokens. Specifically, one must encode the conditioning string into a list of tokens before…