Where Did the Gap Go? Reassessing the Long-Range Graph Benchmark

Benchmark Model Rank Results
node-classification-on-coco-spGPS-tuned#3macro F1: 0.3884±0.0055
node-classification-on-coco-spGatedGCN-tuned#9macro F1: 0.2922±0.0018
node-classification-on-coco-spGINE-tuned#15macro F1: 0.2125±0.0009
node-classification-on-coco-spGCN-tuned#19macro F1: 0.1338±0.0007
node-classification-on-pascalvoc-spGPS-tuned#4macro F1: 0.4440±0.0065
node-classification-on-pascalvoc-spGatedGCN-tuned#8macro F1: 0.3880±0.0040
node-classification-on-pascalvoc-spGINE-tuned#17macro F1: 0.2718±0.0054
node-classification-on-pascalvoc-spGCN-tuned#19macro F1: 0.2078±0.0031
link-prediction-on-pcqm-contactGINE-tuned#11MRR: 0.3509±0.0006MRR-ext-filtered: 0.4617±0.0005
link-prediction-on-pcqm-contactGPS-tuned#12MRR: 0.3498±0.0005MRR-ext-filtered: 0.4703±0.0014
link-prediction-on-pcqm-contactGatedGCN-tuned#13MRR: 0.3495±0.0010MRR-ext-filtered: 0.4670±0.0004
link-prediction-on-pcqm-contactGCN-tuned#15MRR: 0.3424±0.0007MRR-ext-filtered: 0.4526±0.0006
graph-regression-on-peptides-structGCN-tuned#13MAE: 0.2460±0.0007
graph-regression-on-peptides-structGINE-tuned#15MAE: 0.2473±0.0017
graph-regression-on-peptides-structGatedGCN-tuned#17MAE: 0.2477±0.0009
graph-regression-on-peptides-structGPS-tuned#25MAE: 0.2509±0.0014
graph-classification-on-peptides-funcGCN-tuned#22AP: 0.6860±0.0050
graph-classification-on-peptides-funcGatedGCN-tuned#25AP: 0.6765±0.0047
graph-classification-on-peptides-funcGINE-tuned#28AP: 0.6621±0.0067
graph-classification-on-peptides-funcGPS-tuned#32AP: 0.6534±0.0091