Papers10:05Chinese content条件查询下的假设检验:可学习性与交互的价值如果你搞模型评估设计,这篇理论文章用严格上下界讲清楚了交互测试比固定测试省多少查询:最多二次方,不是指数级。#hypothesis testing#adaptive queries#statistical learningaarXiv cs.LG@Zonghuan XuSourceRead laterReadWorth followingUsefulFollow hypothesis testing
Papers11:22Chinese content用范畴论语言描述与构建统计学习模型这篇用范畴论重讲统计学习模型,少见。适合数学背景的人看,能帮你找到跟机器学习研究接头的入口。#statistical learning#category theory#machine learning#math reasoningaarXiv cs.LG@Congwei SongSourceRead laterReadWorth followingUsefulFollow statistical learning
PapersSelected09:22Chinese contentHuman-AI 团队协作的校准视角研究这篇论文为设计更可靠的Human-AI协作系统提供了理论基石,做AI系统设计或人机交互研究的团队值得关注,能帮你理解校准假设如何影响团队决策的可靠性。#human-AI collaboration#calibration#statistical learningaarXiv cs.LG@Eric Nalisnick et al. (4)SourceRead laterReadWorth followingUsefulFollow human-AI collaboration
PapersSelected11:13Chinese content模型崩溃何时发生?结构化交互学习中的条件做生成模型训练或数据增强的团队,这篇论文帮你搞清楚多模型交互时什么时候会踩坑——交互图拓扑是决定因素,值得仔细看看理论条件。#model collapse#generative models#statistical learningaarXiv cs.LG@Yuchen Wu et al. (3)SourceRead laterReadWorth followingUsefulFollow model collapse