Kimi 的做法更新鲜一些，采用了 AlphaGo-Master 的思路，通过提示工程构建的 CoT 轨迹进行轻量级的 SFT 预热。 回想当时在 o1 出现后，无数人想要复现 ...

In 2016, the Go world was rocked after Lee was defeated by AlphaGo, an AI program made by Google's DeepMind. Lee lost 4 out ...