arXiv cs.AI· Weidong Tang, Ruiheng Wang, Jingxuan Zhang, Hexin Wang, Haoyuan Chen, Zhengfeiyang Zhang, Yixing Ma, Zichen Jiang, Ziyu Guo, Haodong Li, Zhengbo Jiao, Zhiqing Cui, Xinxiang Yin, Shuo Liang, Zhenglin Wan, Hongzhan Lin, Wangbo Zhao Kaipeng Zhang, Pengfei Zhou, Odin Zhang, Yang You·· 2 小时前AI 评分41
The Trinity：面向多模态智能体的能力驱动递归自我改进框架
The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents
AI 导读
The Trinity 用 Creator、Explorer、Evolver 三个角色构建递归自我改进框架，把原始任务树重组为分层能力树。其 Capability-Driven Real2Sim2Real 将同节点异构任务抽象为可验证的简化模拟世界，Capability-Aware Harness Router 再按公开任务信息选配 Harness。
来源：arXiv cs.AI · arxiv.org