Paper studies effective trajectories for training terminal agents

Arxiv CS.AI (top papers)

Researchers challenge the assumption that stronger code agents make better teachers for post-training. Using Terminal-Lego, they investigate interaction trajectories to improve terminal agent training.

training

Arxiv CS.AI (top papers)·Sidi YangChaofan TaoJierun ChenTiezheng YuRuoyu WangYuxin JiangYiming DuWendong XuJing XiongTaiqiang WuLifeng ShangXiaohui LiNgai WongHaoli Bai

8 days ago