@voooooogel imaging what happens once the whole training corpus is meticulously refined!my impression is that pretraining now is largely raw data, and that's why a lot of heavy handed alignment is necessarybut as those datasets are continually refined, less alignment is needed later
in reply to: 1873133153650655658
Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.