<ul><li>The article introduces GRAFF-LP, an extension of GRAFF for link prediction in heterophilic datasets.</li><li>GRAFF-LP discriminates existing from non-existing edges by implicitly learning to separate the edge gradients.</li><li>A new readout function inspired by physics is proposed, improving performance of GRAFF-LP and other baseline models.</li><li>Heterophily measures specifically tailored for link prediction are suggested, different from those used in node classification.</li></ul>

Link Prediction with Physics-Inspired Graph Neural Networks

Discover more