This letter presents a novel detection strategy for Spatially-Multiplexed Generalized Spatial Modulation systems. It is a multi-stage detection that produces a list of candidates of the transmitted signal vector, sorted according to the proximity of the data vector to one of the possible vector subspaces. The quality metric and list-length metric selects the best candidate and manages the list length, respectively. Performance results show that it significantly reduces the performance gap to the optimal maximum likelihood detector, while maintaining significant computational cost reduction.