KMeans classification applied to the planet layer thickness fractions derived from ExoMDN for the 2025 sample. (A) Classification performed with two assumed populations. (B) Classification with three assumed populations. The fitting algorithm was executed 100 times and stacked on top of each other. For each run, a new sample is made by randomly drawing a new composition for each planet from its four-layer posterior distribution, such that the total thickness of a planet adds up to 1. KMeans then clusters each drawn sample, and the planets are coloured according to their assigned cluster. The top cluster (cluster in which the mean density of all member planets is the highest) is coloured red, the bottom cluster is coloured yellow, and the middle (for three populations) is coloured blue. Planets that do not match the three primary colours have been assigned to different populations in various iterations.