00:00:30 给AI一支笔，它能撬动地球吗？ 

00:04:45 给AI请个“外援”，裁判才能更靠谱 

00:09:09 和AI说话的艺术：你以为是聊天，其实是盖楼 

00:14:29 AI界的“庖丁解牛”：如何用“乐高积木”搞定复杂工程难题？ 

本期介绍的四篇论文：
[LG] Thinking Isn't an Illusion: Overcoming the Limitations of Reasoning Models via Tool Augmentations  

[UC Berkeley & Northeastern University]  

https://arxiv.org/abs/2507.17699  

---

[CL] Can External Validation Tools Improve Annotation Quality for LLM-as-a-Judge?  

[University of Cambridge & Apple]  

https://arxiv.org/abs/2507.17015  

---

[LG] Understanding Prompt Programming Tasks and Questions  

[CMU & JetBrains Research]  

https://arxiv.org/abs/2507.17264  

---

[LG] A Learning-based Domain Decomposition Method  

[University of Cambridge & NVIDIA]  

https://arxiv.org/abs/2507.17328