你有没有想过，AI除了会聊天画画，还能做什么更酷的事？本期节目，我们将一口气看到AI能力的多个惊人侧面。从像人一样“脑补”物理世界，到用“笨方法”实现更高效的学习，再到成为物理学家的“科研搭子”，解决真正的科学难题。这些最新论文将刷新你对AI潜力的认知！

00:00:28 AI学会了“脑补”，世界就大不一样了

00:06:08 大模型里的“关系户”，它凭什么吸引了所有注意力？

00:13:10 AI省钱的终极奥义，少就是多

00:18:04 一个“笨方法”，让AI学得更快

00:22:49 AI，从聊天高手到科研搭子

本期介绍的几篇论文：

[LG] Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling  

[CMU & UT Austin & Brown University]  

https://arxiv.org/abs/2603.04553 

---

[CL] The Spike, the Sparse and the Sink: Anatomy of Massive Activations and Attention Sinks  

[New York University]  

https://arxiv.org/abs/2603.05498 

---

[CL] Sparse-BitNet: 1.58-bit LLMs are Naturally Friendly to Semi-Structured Sparsity  

[Microsoft Research]  

https://arxiv.org/abs/2603.05168 

---

[CL] Replaying pre-training data improves fine-tuning  

[Stanford University]  

https://arxiv.org/abs/2603.04964 

---

[AI] Solving an Open Problem in Theoretical Physics using AI-Assisted Discovery  

[Google Research]  

https://arxiv.org/abs/2603.04735