A arxiv.org Digital Agent outperforms o1 by 15% – trained with new RL-variant similar to R1 11 points let_tim_cook_ 1 year ago 0 comments
