Avatarl: Training langauge models from scratch with pure RL

2 points| krkartikay | 6 months ago |tokenbender.com