[人人能懂] 从死记硬背到举一反三的五堂课

00:00:27 高手过招：抄作业还是自己闯？00:04:38 AI也会“喜新厌旧”？高手是如何做到“学而不忘”的00:09:36 让AI学会“脑补”，速度提升5倍的秘密00:13:45 AI的“开窍”秘诀：如何让机器学会举一反三？00:17:53 聪明反被聪明误？AI的“笨办法”本期介绍的五篇论文：[LG] Towards a Unified View of Large Language Model Post-Training  [Tsinghua University]  https://arxiv.org/abs/2509.04419 ---[LG] RL's Razor: Why Online Reinforcement Learning Forgets Less  [MIT]  https://arxiv.org/abs/2509.04259 ---[LG] Set Block Decoding is a Language Model Inference Accelerator  [FAIR at Meta]  https://arxiv.org/abs/2509.04185 ---[LG] ArcMemo: Abstract Reasoning Composition with Lifelong LLM Memory  [University of California, San Diego]  https://arxiv.org/abs/2509.04439 ---[LG] Learning When to Plan: Efficiently Allocating Test-Time Compute for LLM Agents  [University College London & University of Oxford]  https://arxiv.org/abs/2509.03581 在小宇宙查看该单集文稿