3 ms·
In the paper it is mentioned that the LLMs predicting the previous token are indeed pre-trained in this way, so it is not true that the difference is obvious.
by frotaur 3y ago
In the paper it is mentioned that the LLMs predicting the previous token are indeed pre-trained in this way, so it is not true that the difference is obvious.