arXiv · NLP· NeoHorse Team, Guoliang Cao, Guohao Dai, Tianyu Guo, Kai Han, Hailin Hu, Zihan Jiang, Xiang Kuang, Boxun Li, Yulong Li, Zehua Pei, Yuchuan Tian, Jiamin Wang, Yu Wang, Yunhe Wang, Yihong Wu, Haiyang Xu, Shuo Zhang, Hang Zhou, Siyang Cheng, Jiayu Fan, Wei He, Qingrui Jiao, Hongguang Li, Zhiyuan Li, Runke Liu, Xi Liu, Xinchen Liu, Sinno Jialin Pan, Yi Ren, Liuyang Song, Chenyu Wang, Bei Yu, Quanlu Zhang, Xiangyu Zhang, Mengyu Zheng, Yingjie Zong·· 2026-10-16AI 评分57
NeoHorse-1: 通过 Agentic 后训练和路由架构探索递归自我改进
NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness
AI 导读
NeoHorse-1 是一个探索递归自我改进(RSI)的 agent 原生模型家族,通过智能路由记录交互并转化为训练样本,包含结构验证和多维语义评估。该方法将评估反馈转化为下一轮训练混合,在 11 个基准测试上使 4B 模型宏平均准确率从 58.94 提升至 64.87,9B 模型从 65.60 提升至 69.04,显著缩小了后训练 4B 模型与 9B 基座模型的差距。
来源:arXiv · NLP · arxiv.org