你是否感觉努力到一定阶段就会遭遇瓶颈，或者在海量信息中难以抉择？本期我们要聊的几篇最新论文，就揭示了AI如何用更聪明的策略解决这些难题。我们将看到，机器不仅学会了在优化的“峡谷”中奋力一跃，还掌握了从招聘到投资都通用的决策“万能公式”。更有趣的是，AI已经开始懂得“吃一堑长一智”，甚至能在训练的最后一百米，像顶尖运动员一样进行科学冲刺。00:00:33 从龟速到全速，如何跳出优化的“平庸峡谷”00:05:48 决策的万能公式00:11:30 高手过招，如何在信息爆炸中做出聪明选择？00:17:42 为什么你家的扫地机器人总像个“笨蛋”？00:24:16 AI冲刺的最后一百米，怎么跑才最快？本期介绍的几篇论文：[LG] Local linear convergence of gradient methods for overparameterized Gaussian mixtures [University of Washington] https://arxiv.org/abs/2605.30936 ---[LG] Universal Decision Learners [Adobe Research] https://arxiv.org/abs/2605.30694 ---[LG] Hedging on the Frontier: Learning New Tasks with Few Samples [ETH Zurich & UC San Diego] https://arxiv.org/abs/2605.30997 ---[RO] Don't Fool Me Twice: Adapting to Adversity in the Wild with Experience-Driven Reasoning [Indian Institute of Technology & CMU] https://arxiv.org/abs/2605.31119 ---[CL] Towards Efficient LLMs Annealing with Principled Sample Selection [Microsoft Research Asia] https://arxiv.org/abs/2605.31175 在小宇宙查看该单集文稿

Ask this episode

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

Transcript

Related