← back to paper
arxiv: 2606.17882 · 2 revisions
Structural Preservation and the Logical Expressiveness of Graph Neural Networks