00:00:33 高手与普通人的差距，在于“记忆预算”的分配

00:04:16 AI当“牛顿”：我们如何找到万物生长的公式？ 

00:08:26 让AI不止听话，更要会提问 

00:12:09 AI 思考的艺术：如何做到又快又好？ 

00:18:06 AI识人心：20盘棋，就“看穿”了你 

本期介绍的五篇论文：

[LG] Capacity-Constrained Continual Learning  

[Google DeepMind]  

https://arxiv.org/abs/2507.21479  

---

[LG] EvoSLD: Automated Neural Scaling Law Discovery With Large Language Models  

[Peking University & Tsinghua University]  

https://arxiv.org/abs/2507.21184  

---

[LG] Teaching Language Models To Gather Information Proactively  

[Microsoft]  

https://arxiv.org/abs/2507.21389  

---

[LG] TriangleMix: A Lossless and Efficient Attention Pattern for Long Context Prefilling  

[Microsoft Research]  

https://arxiv.org/abs/2507.21526  

---

[LG] Learning to Imitate with Less: Efficient Individual Behavior Modeling in Chess  

[University of Toronto]  

https://arxiv.org/abs/2507.21488