[人人能懂] 驯服AI的终极武器，藏在40年前的数学和一张图里

00:00:36 AI“减肥”的秘密，被一个40年前的算法破解了

00:05:46 驯服AI的终极武器：不是模糊奖励，而是清晰清单

00:09:56 AI的“开窍”秘诀：如何教机器像高手一样思考？ 

00:14:53 给AI请个好私教的秘密 

00:19:50 给AI画张图，它能看得更明白吗？

本期介绍的五篇论文：

[LG] The Geometry of LLM Quantization: GPTQ as Babai's Nearest Plane Algorithm  

[ISTA & ETH Zürich]  

arxiv.org 

---

[CL] Checklists Are Better Than Reward Models For Aligning Language Models  

[CMU & Apple]  

arxiv.org 

---

[LG] Revisiting LLM Reasoning via Information Bottleneck  

[The University of Sydney & ByteDance]  

arxiv.org 

---

[CL] TRPrompt: Bootstrapping Query-Aware Prompt Optimization from Textual Rewards  

[EPFL]  

arxiv.org 

---

[LG] Does visualization help AI understand data?  

[Harvard University]  

arxiv.org