[人人能懂] 从模仿到洞察：AI如何学会像侦探一样思考？

00:00:34 你以为AI是高科技？它的底层逻辑可能几亿年前就有了

00:05:07 秒懂的秘密：如何看透行为背后的“算法”

00:09:28 AI大模型里的“快递小哥”：不只求快，还要求稳

00:14:22 AI的“双档切换”：如何用最聪明的大脑，干最划算的活？

00:18:24 我们如何思考“如果”：一种全新的思维工具

本期介绍的五篇论文：

[LG] Understanding Transformers through the Lens of Pavlovian Conditioning  

[Meta Platforms, Inc]  

https://arxiv.org/abs/2508.08289  

---

[RO] Rational Inverse Reasoning  

[MIT CSAIL]  

https://arxiv.org/abs/2508.08983  

---

[LG] Scaled-Dot-Product Attention as One-Sided Entropic Optimal Transport  

[Stanford University]  

https://arxiv.org/abs/2508.08369  

---

[LG] OverFill: Two-Stage Models for Efficient Language Model Decoding  

[Cornell University]  

https://arxiv.org/abs/2508.08446  

---

[LG] Topos Causal Models  

[Adobe Research]  

https://arxiv.org/abs/2508.08295