Enormous experimental and theoretical studies have been conducted on the order-to-disorder transition (ODT) of block copolymers (BCPs). Thanks to the earlier random phase approximation (RPA) and self-consistent field theory (SCFT) calculations, it is now well known that the ODT point of symmetric AB BCP melt is (χN)ODT = 10.495 in the mean field limit, when standard Gaussian chain model is adopted. However, the story becomes more complicated for short polymer chains for which Gaussian chain model is not appropriate and finite-range interaction is naturally present in the model, even before considering the compositional fluctuation effect. In this talk, we first present RPA calculation result of discrete chain model with finite-range interaction and show that the ODT change strongly depends on the chain model. In a series of publications on ODT change of thin BCP films, we utilized SCFT calculation with finite-range interaction to confirm the experimentally observed ODT change when the interfaces are preferential and/or neutral. We also performed Langevin field-theoretic simulation boosted by deep learning technique in an effort to identify the effect of compositional fluctuation at experimentally relevant low invariant polymerization index regime.