你有没有想过，我们不仅能使用AI，还能看透它的“内心世界”吗？本期，我们将跟随几篇最新论文，扮演一次AI的建筑师、心理医生和侦探，看看如何重塑一个更省钱的推荐系统，如何看穿AI从无害信息中“悟”出的危险想法，以及如何用一把“手术刀”解剖并教会它举一反三。让我们一起出发！

00:00:28 推荐系统的大难题：如何让它更聪明，还更省钱？

00:05:59 你教AI画苹果，它却学会了当海盗

00:14:26 我们终于有了一把解剖AI的“手术刀”

00:20:44 AI侦探：你的模型到底跟谁学的？

00:27:06 教AI举一反三：为什么聪明的模型，一到考场就蒙圈？

本期介绍的几篇论文：

[IR] Meta Lattice: Model Space Redesign for Cost-Effective Industry-Scale Ads Recommendations

[Meta AI]

https://arxiv.org/abs/2512.09200

---

[CL] Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs

[Truthful AI]

https://arxiv.org/abs/2512.09742

---

[LG] Provably Learning from Modern Language Models via Low Logit Rank

[Microsoft Research & UC Berkeley & MIT]

https://arxiv.org/abs/2512.09892

---

[LG] Natural Geometry of Robust Data Attribution: From Convex Models to Deep Networks

[The University of Texas at Austin]

https://arxiv.org/abs/2512.09103

---

[LG] Closing the Train-Test Gap in World Models for Gradient-Based Planning

[Columbia University]

https://arxiv.org/abs/2512.09929

[人人能懂] 从系统蓝图到思维手术刀

AI可可AI生活