PapersBuilders10:53Chinese content时间序列基础模型预测准不代表因果效应估计准做定价或销量分析的朋友值得看看:论文用真实零售数据证明,TabPFN 这类基础模型预测准但因果效应估计可能全错,连工具变量用错形式还会更糟。需要从预测走向因果决策的团队建议读一下。#time series#causal inference#foundation models#instrumental variablesaarXiv cs.LG@Juwon Hong et al. (3)SourceRead laterReadWorth followingUsefulFollow time series
PapersBuilders09:51Chinese contentCAT-CCB:面向未测混杂的离线强化学习迁移框架离线 RL 遇上混杂变量和域差异怎么办?这篇用工具变量加负控制给出可识别方案,做因果 RL 的可以看看推导。#offline reinforcement learning#causal inference#instrumental variables#CAT-CCBaarXiv cs.LG@Junyi Liao et al. (3)SourceRead laterReadWorth followingUsefulFollow offline reinforcement learning
ModelsSelected11:11Chinese contentA-IHF: 图扩散残差提取器用于控制函数工具变量新方法A-IHF搞定了工具变量残差提取#A-IHF#instrumental variables#causal inferenceaarXiv cs.LG@Rui Wu et al. (5)SourceRead laterReadWorth followingUsefulFollow A-IHF