Graph Set Transformer: Learning Over Sets of Graphs

arXiv 2606.05116·medium signal

Escrig Molina, Chen, and Probst introduce the Graph Set Transformer (GST), an architecture for tasks where each example is a set of graphs rather than a single graph. It targets per-set prediction problems underserved by standard GNNs. Of interest to practitioners in cheminformatics, molecular ML, and structured-data domains.