Avatarl: Training language models from scratch with pure reinforcement learningtokenbender.com2 points·haneefmubarak··0 commentsOpen articleSaveView on HN