What Expressivity Theory Misses: Message Passing Complexity for GNNs

NeurIPSSpotlight2025

Authors
Niklas Kemper, Tom Wollschläger, Stephan Günnemann
Venue
NeurIPS 2025
Track
Spotlight

TL;DR

Expressivity theory, characterizing which graphs a GNN can distinguish, has become the predominant framework for analyzing GNNs, with new models striving for higher expressivity…

Opening excerpt from the authors’ abstract. source

Read the paper

Topics

theory graph

← All NeurIPS 2025 Spotlight papers · Browse the whole archive