{"source":{"id":"2112107486098473471","title":"在单题测试里拿第一的决策模型，丢进真实智能体，成功率只有69.5%。 专训智能体状态的4B小模型，却跑出了91.5%。…","url":"https://mubeitech.com/p/2112107486098473471"},"method":"semantic-similarity","count":0,"items":[]}