你有没有想过，最聪明的AI不仅要会解题，更要懂得如何省钱、如何团队协作、甚至如何避免“摸鱼”吗？本期节目，我们将一口气解读几篇最新论文，看看AI如何通过精细的“拆解”来降本增效，如何组建“AI教练天团”实现自我进化，又是如何学会“察言观色”，从一个笨拙的工具，变身为高情商的队友。准备好了吗？让我们一起揭开AI“人情世故”的秘密。00:00:32 为什么你的AI服务又贵又慢？答案藏在一个“拆”字里00:05:48 AI当教练，一句话教会AI当车神00:11:58 AI学会了“省钱”，这对我们有什么启发？00:17:43 AI也会摸-鱼？一个团队的智慧，是怎么被“猪队友”拖垮的00:23:55 如何让你的AI助理，从“笨蛋”变“高情商”？本期介绍的几篇论文：[LG] From Models to Operators: Rethinking Autoscaling Granularity for Large Generative Models[Rice University & Microsoft Research]https://arxiv.org/abs/2511.02248---[LG] Automated Reward Design for Gran Turismo[University of Montreal & Turing Inc. & Sony AI]https://arxiv.org/abs/2511.02094---[LG] Re-FORC: Adaptive Reward Prediction for Efficient Chain-of-Thought Reasoning[AWS Agentic AI]https://arxiv.org/abs/2511.02130---[LG] Unlocking the Power of Multi-Agent LLM for Reasoning: From Lazy Agents to Deliberation[The Pennsylvania State University & Harvard University & Michigan State University]https://arxiv.org/abs/2511.02303---[LG] Training Proactive and Personalized LLM Agents[CMU]https://arxiv.org/abs/2511.02208

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related