high temp often gets (ab)used to "make models more creative," but it's really a hack, because logprobs conflate semantic unlikelihood with syntactic unlikelihood--on 405-base, "peeling the the( …)" is many OOMs more likely than "peeling the demonic (fruit)"i think the future… https://t.co/wW2dBLO2gt
cited on: llama-3-1-405b-base
Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.