 This article compares mean field models for predicting grain size distribution evolution under thermomechanical solicitations and discusses different microstructure representations, including topology in neighborhood construction. Experimental data from a heat treatment campaign on 316L austenitic stainless steel are used to identify material parameters and compare model predictions. The study shows that introducing neighborhood topology can improve predictions in monomodal cases, but has only small improvements in terms of implementation complexity compared to classical mean field models. This article was authored by Marian Roth, Baptiste Flippen, Natalie Bezolo, and others.