We introduce a class of comparison-based preference learning models that extend (Generalized) Bradley-Terry, can generalize, and have monotonicity guaranties.