[arXiv]score: 0.52
FSpecGNNs Extend Spectral GNNs Beyond 1-WL Expressivity via Bivariate Filters
May 26, 2026
Full-Spectrum GNNs lift spectral filtering to node-pair domains with bivariate eigenvalue filters, achieving expressivity up to Local 2-GNN while universally approximating node-pair signals for heterophilic graph learning.
cs.LG
HOW THIS AFFECTS YOU
●
researcherThis provides a theoretically grounded path past the 1-WL expressivity ceiling for spectral GNNs, with proven universality for node-pair signals relevant to heterophilic settings.