一、论断写在前面论文题目，Smaller，Weaker，YetBetter，TrainingLLMReasonersviaCompute，OptimalSampling论文链接，​​https，arx...