Experiments that compare or train with both parts are collected here.
Joint residual GNN, held-out runs and all-model ranking.
Open resultsNode sizes, shape fidelity and label transfer.
Open resultsAll preprocessing, models and visual inspections.
Open results