这份149页的报告把长周期智能体的进化路径说清楚了,还给出了难度和能力分层,做智能体的人值得一看。
中国人民大学GAIR联合多机构发布149页长周期智能体调查报告,提出H1-H3任务难度层级和C1-C3能力层级。研究发现任务跨度每4-7个月翻倍,并给出两条进化主线(Harness Engineering和Model Optimization)。报告为智能体长期规划能力提供了系统框架。
149 Pages Mapping the Long-Horizon Agent Frontier: Multi-University Survey Proposes Harness Engineering and Model Optimization as Two Main Evolution Lines for Next-Generation AI Agents
Renmin University GAIR leads multi-institution 149-page survey on long-horizon agents, proposing H1-H3 task difficulty hierarchy and C1-C3 capability tiers, with task span doubling every 4-7 months.