Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Weisfeiler-Leman Is Incomplete on Simple Spectrum Graphs, so Canonicalize Them

About

Graphs with a simple spectrum admit cubic-time isomorphism testing, yet we prove that for every natural number $k$, the $k$-Weisfeiler-Leman ($k$-WL) test cannot distinguish all non-isomorphic graphs with a simple spectrum. As the WL hierarchy upper-bounds the distinguishing power of widely-used Graph Neural Networks (GNNs), this incompleteness applies to all such GNNs, ruling out completeness for every $k$-WL-aligned GNN family. To close this gap, we introduce PRiSM (Partition, Refine, Solve, Match), the first provably complete canonicalization of simple-spectrum eigendecompositions. PRiSM obtains the completeness guarantee that prior canonicalizations provably lack, and resolves the open problem of achieving complete expressivity on simple-spectrum graphs. When composed with DeepSets or a Transformer, PRiSM achieves universal approximation on simple-spectrum graphs, justifying the use of canonicalized Laplacian positional encodings. Empirically, PRiSM performs comparably to or outperforms existing spectral canonicalizations on graph regression, classification, and expressivity

Snir Hordan, Nadav Dym, Tim Seppelt• 2026

Related benchmarks

TaskDatasetResultRank
Molecular property predictionMolHIV
ROC-AUC74.3
39
Molecular property predictionMOLTOX21
ROC-AUC0.783
33
Graph ExpressivityBREC
Basic Expressivity Score60
26
Molecular property predictionZINC-12K
MAE0.103
8
Molecular property predictionOGB MOLPCBA
AP27.8
7
Molecular property predictionOGB molclintox
ROC-AUC85.76
4
Molecular property predictionOGB MOLBBBP
ROC-AUC0.7088
4
Molecular property predictionalchemy
MAE0.1196
4
Showing 8 of 8 rows

Other info

Follow for update