Abstract: Subset selection has been widely studied but remains underexplored for synthetic tabular data, particularly in data sharing contexts that require high quality data. While generative models ...