someone should train a model to predict the "previous" token by just reversing all the training data

Posters: Shoalst0ne. Shoalst0ne

7 likes recorded by the source corpus.