We describe a fully automatized homogenization procedure and illustrate it on Argentinean weather station temperature series. The procedure relies on multiple pairwise comparisons between a candidate station and its surrounding stations. The main advantage of this approach is to get around the difficulty of defining a reliable reference series; its main drawback is to often require visual attribution and grouping of shifts resulting in too high a cost in human time for implementation on large datasets. Here, we fully automatize these two steps by using a probabilistic metric of similarity between shifts which is leveraged within two optimized clustering schemes. Simulation results show performance improvements versus both visual inspection and the automatized procedure of Menne MJ, Williams CN, Jr. 2009. Homogenization of temperature series via pairwise comparisons. J. Clim. 22: 1700–1717. Implementation on Argentinean temperature series results in the identification and removal of numerous inhomogeneities; corrected series reveal stronger and spatially smoother warming trends.