Pure Transformers are Powerful Graph Learners

Benchmark Model Rank Results
graph-classification-on-ddTokenGT#45Accuracy: 73.950±3.361
graph-classification-on-imdb-bTokenGT#6Accuracy: 80.250±3.304
graph-classification-on-nci1TokenGT#42Accuracy: 76.740±2.054
graph-classification-on-nci109TokenGT#28Accuracy: 72.077±1.883
graph-regression-on-esr2TokenGT#8R2: 0.641±0.000RMSE: 0.529±0.641
graph-regression-on-f2TokenGT#8R2: 0.872±0.000RMSE: 0.363±0.872
graph-regression-on-kitTokenGT#8R2: 0.800±0.000RMSE: 0.486±0.800
graph-regression-on-lipophilicityTokenGT#16RMSE: 0.852±0.023R2: 0.545±0.024
graph-regression-on-parp1TokenGT#8R2: 0.907±0.000RMSE: 0.383±0.907
graph-regression-on-pcqm4mv2-lscTokenGT#17Validation MAE: 0.0910Test MAE: 0.0919
graph-regression-on-peptides-structTokenGT#20MAE: 0.2489±0.0013
graph-regression-on-pgrTokenGT#5R2: 0.684±0.000RMSE: 0.543±0.684
graph-regression-on-zinc-fullTokenGT#13Test MAE: 0.047±0.010
molecular-property-prediction-on-esolTokenGT#12RMSE: 0.667±0.103R2: 0.892±0.036
molecular-property-prediction-on-freesolvTokenGT#8RMSE: 1.038±0.125R2: 0.930±0.018