跳到正文
arXiv · ML· LoongSage Team, Bowei Song, Caiwei Zhao, Chuwen Liu, Dechuang Gu, Dong Fang, Feiyu Xie, Huaiyu Li, Hui Zhang, Jian Tian, Jian Wang, Jie Zhao, Liguang Li, Maoyun Wang, Meiyan Yang, Minghan Zhu, Pai Liu, Pandeng Yao, Renbao Tian, Ru Yang, Shuai Jia, Shuai Tan, Shuailong Li, Shuliang Li, Teng Jin, Tianze Yang, Wenbo Cui, Wenjing Bai, Wenkang Wu, Xiaowei Jiang, Xiaoyi Dong, Yanpeng Wang, Yanwen Wang, Yaohui Ma, Yawen Wang, Yihang Zhang, Yingjie Yan, Zhanbiao Lian, Zhenyu Lu, Zhida Hu, Zhihui Hao, Zhihui Hu, Zhiwen Xiao, Zhiyi Gao, Zhongwei Qi, Zhuangzhuang Wu, Ziqian Liu, Ziyue Zheng·· 2026-10-14AI 评分43

LoongSage:面向前沿大模型的高效智能体强化学习框架

LoongSage: A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes

AI 导读

LoongSage 是一个全异步的智能体原生强化学习框架,核心代码约 1.4 万行,基于 Megatron 和 SGLang 构建。 该框架引入 AgentFlow 范式协调白盒与黑盒智能体,并利用滑动窗口策略限制轨迹过时性,同时通过 TransferMesh 机制实现拓扑感知的同步。 四项完整训练评估表明,该全异步流程消除了同步空闲气泡,在提升任务奖励的同时维持了硬件利用率。

来源:arXiv · ML · arxiv.org