Tag
Percy Liang announces that for the next Marin model, they are compiling a new data mix and request high-quality token data for pre-training, mid-training, and SFT.