Monotone and Separable Set Functions: Mathematical Foundations and Neural Models

A recent preprint on arXiv (2510.23634) presents Monotone and Separating (MAS) set functions, which maintain the inherent partial order of sets: S ⊆ T if and only if F(S) ≤ F(T). The study defines both lower and upper limits on the vector dimension necessary for achieving MAS functions, influenced by the size of multisets and the foundational ground set. It is noted that MAS functions cannot exist for infinite ground sets; however, a modified 'weakly MAS' model is introduced, demonstrating stability through Hölder continuity. Additionally, the authors reveal that MAS functions can create universally monotone models and can approximate all monotone set functions, although the abstract provides limited details on experimental findings.