CauTabBench

CauTabBench是由KTH皇家理工学院开发的用于评估表格数据合成模型的高阶结构因果基准框架。该数据集旨在通过随机抽样的因果图生成基准数据集，用于训练和评估表格合成模型。数据集的创建过程涉及使用因果发现方法从基准和合成数据集中提取因果信息，并应用高阶结构因果度量进行比较。CauTabBench主要应用于解决表格数据合成中的复杂依赖捕获问题，特别是在分布偏移下的预测、自动化决策制定和跨表格理解等下游任务中。CauTabBench is a high-order structural causal benchmark framework developed by KTH Royal Institute of Technology for evaluating tabular data synthesis models. It is designed to generate benchmark datasets via randomly sampled causal graphs for training and evaluating tabular synthesis models. The dataset creation process involves extracting causal information from both benchmark and synthetic datasets using causal discovery methods, and conducting comparisons via high-order structural causal metrics. CauTabBench is primarily applied to address the challenge of capturing complex dependencies in tabular data synthesis, especially for downstream tasks including prediction under distribution shift, automated decision-making, and cross-table understanding.

搜集汇总

背景与挑战

背景概述

CauTabBench是由KTH皇家理工学院开发的高阶结构因果基准框架，旨在通过随机抽样的因果图生成基准数据集，用于评估表格数据合成模型的性能。该数据集利用因果发现方法提取因果信息，并应用高阶结构因果度量进行比较，主要应用于解决表格数据合成中的复杂依赖捕获问题，特别是在分布偏移下的预测和自动化决策等下游任务中。