Scaling interactive environments is important for training long-horizon agents through reinforcement learning. These environments often require further testi…
机构:复旦
来源:arXiv 2610.18067 | AI4Papers 论文推荐平台