4B小模型数学推理首超Claude 4，700步RL训练逼近235B性能 | 港大&字节Seed&复旦 - AI工具导航AI Tool Navigation

4B小模型数学推理首超Claude 4，700步RL训练逼近235B性能 | 港大&字节Seed&复旦

Direct Link ">View on Mobile

Content assisted by AI. Please review carefully.

Related Navigation