Rather than performing conformal aggregation in the space of prediction regions, we propose a vector extension of the conformal score, through which aggregation can be performed more efficiently.