Long-horizon agentic tasks demand strong reasoning and efficient execution across successive interactions with dynamic environments. A common approach decoup…
机构:南京大学
来源:arXiv 2609.38891 | AI4Papers 论文推荐平台