4月7日,阿里通义实验室智能计算团队宣布推出新算法FIPO(Future-KL Influenced Policy Optimization),引入Future-KL机制,奖励关键Token,解决纯强化学习(Pure RL)训练中“推理长度停滞”难题。据该团队介绍,在32B规模的纯RL设定下,率先实现对o1-mini与同规模DeepSeek-Zero-MATH的性能反超。
特别声明:以上内容(如有图片或视频亦包括在内)为自媒体平台“网易号”用户上传并发布,本平台仅提供信息存储服务。
Notice: The content above (including the pictures and videos if any) is uploaded and posted by a user of NetEase Hao, which is a social media platform and only provides information storage services.