[人人能懂] 从“自学成才”到“组团思考”的AI新范式

00:00:28 AI的“学霸”和“教练”，如何合二为一？00:04:33 想让AI更聪明？关键不是砸钱，是“会花钱”00:08:26 AI的“朋友圈”：单个脑补，不如组团思考00:13:12 如何让AI既是“通才”，又是“专才”？00:16:07 数据世界的“蝴蝶效应”：我们如何揪出那个扇动翅膀的“坏数据”？本期介绍的五篇论文：[LG] Your Reward Function for RL is Your Best PRM for Search: Unifying RL and Search-Based TTS  [Rutgers University & Nanyang Technological University]  https://arxiv.org/abs/2508.14313  ---[LG] Compute-Optimal Scaling for Value-Based Deep RL  [UC Berkeley]  https://arxiv.org/abs/2508.14881  ---[LG] Graph Concept Bottleneck Models  [Stony Brook University & University of California, San Diego & IBM Research]  https://arxiv.org/abs/2508.14255  ---[LG] Amortized Bayesian Meta-Learning for Low-Rank Adaptation of Large Language Models  [Princeton University]  https://arxiv.org/abs/2508.14285  ---[LG] Understanding Data Influence with Differential Approximation  [University of Hong Kong & Chinese University of Hong Kong]  https://arxiv.org/abs/2508.14648  在小宇宙查看该单集文稿