今天我们不聊模型参数有多大，而是聊如何让AI变得更“会思考”，这种思考方式，有时甚至有些反常识。比如，为什么给AI疯狂“补课”，它反而可能越学越笨？我们还会探讨，如何像一位高明的老师一样引导AI攻克难题，而不是直接灌输答案。更进一步，我们会揭示如何训练AI像个侦探一样，学会“讲道理”地分析代码，以及如何让整个系统学会动态协作，找到最高效的“偷懒”方式。

00:00:35 AI大模型时代，如何花小钱办大事？

00:05:47 给AI“补课”的陷阱，为什么学得越多，它反而越笨？

00:11:37 高手辅导功课，为什么不直接给答案？

00:16:48 让AI学会“讲道理”，代码世界的侦探是怎样炼成的？

00:22:00 让AI学会“省时间”，一种更聪明的快

本期介绍的几篇论文：

[LG] Rich Insights from Cheap Signals: Efficient Evaluations via Tensor Factorization  

[Google DeepMind & University of Michigan]  

https://arxiv.org/abs/2603.02029 

---

[LG] Theoretical Perspectives on Data Quality and Synergistic Effects in Pre- and Post-Training Reasoning Models  

[University of Southern California & University of California Los Angeles & Google Research]  

https://arxiv.org/abs/2603.01293 

---

[LG] Learn Hard Problems During RL with Reference Guided Fine-tuning  

[ByteDance Seed & UC Berkeley & CMU]  

https://arxiv.org/abs/2603.01223 

---

[LG] Agentic Code Reasoning  

[Meta]  

https://arxiv.org/abs/2603.01896 

---

[CL] Learning to Draft: Adaptive Speculative Decoding with Reinforcement Learning  

[Microsoft Research Asia & Peking University]  

https://arxiv.org/abs/2603.01639