Graph Neural Networks Can (Often) Count Substructures

Paolo Pellizzoni, Till Hendrik Schulz, Karsten M. Borgwardt
2/12/2026

No official code found. Best implementation: BorgwardtLab/GNNsCanCountSubstructures (5 stars, MIT, last commit 2025-02-27).

Semantic Scholar

Code Implementations(1)

BorgwardtLab/GNNsCanCountSubstructuresCommunity69%
MIT

Code and Data for the paper: "Graph Neural Networks Can (Often) Count Substructures" (ICLR 2025 Spotlight)

50GLSL, ShellFeb 27, 20251 years agoMIT
gnnsiclr2025subgraph-isomorphism

Cite this paper

@article{pellizzoni2026graph,
  title  = {Graph Neural Networks Can (Often) Count Substructures},
  author = {Paolo Pellizzoni and Till Hendrik Schulz and Karsten M. Borgwardt},
  year   = {2026},
  url    = {https://api.semanticscholar.org/CorpusID:be8e2c90dd878ad4f7f6588f701bd7bede1a57ba},
  journal = {ICLR 2025 2025}
}

Discussion