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
정보
- 프로그램
- 주기매일 업데이트
- 발행일2025년 11월 6일 오전 1:09 UTC
- 길이30분
- 등급전체 연령 사용가
