Topological Deep Learning is hard!

One of the first papers with AIDOS, we find strong evidence that many message passing based models are overffitting on the combinatorics. More on this at No Triangulation Without Representation: Generalization in Topological Deep Learning