| Transformers Meet Directed Graphs | ✓ Link | 0.2222 ± 0.0010 | No | 0.2044 ± 0.0020 | 14378069 | SAT++ with Magnetic Laplacian | 2023-01-31 |
| []() | | 0.2222 ± 0.0032 | No | 0.2044 ± 0.0020 | 14378069 | SAT++ with Magnetic Laplacian | |
| Transformers over Directed Acyclic Graphs | ✓ Link | 0.2018 ± 0.0021 | No | 0.1846 ± 0.0010 | 14952882 | DAGformer | 2022-10-24 |
| Structure-Aware Transformer for Graph Representation Learning | ✓ Link | 0.1937 ± 0.0028 | No | 0.1773 ± 0.0023 | 15734000 | SAT | 2022-02-07 |
| Unlocking the Potential of Classic GNNs for Graph-level Tasks: Simple Architectures Meet Excellence | ✓ Link | 0.1896 ± 0.0024 | | 0.1742 ± 0.0027 | | GatedGCN+ | 2025-02-13 |
| Recipe for a General, Powerful, Scalable Graph Transformer | ✓ Link | 0.1894 | No | 0.1739 ± 0.001 | 12454066 | GPS | 2022-05-25 |
| []() | | 0.1830 ± 0.0024 | No | 0.1661 ± 0.0012 | 9053246 | GraphTrans (GCN-Virtual) | |
| []() | | 0.1770 ± 0.0012 | No | 0.1631 ± 0.0090 | 63684290 | GMAN+bag of tricks | |
| []() | | 0.1751 ± 0.0049 | No | 0.1607 ± 0.0040 | 35246814 | DAGNN | |
| Directed Acyclic Graph Neural Networks | ✓ Link | 0.1751 ± 0.0049 | | 0.1607 ± 0.0040 | | DAGNN | 2021-01-20 |
| []() | | 0.1751 ± 0.0015 | No | 0.1599 ± 0.0009 | 7563746 | GraphTrans (GCN) | |
| Do We Need Anisotropic Graph Neural Networks? | ✓ Link | 0.1595 ± 0.0019 | No | 0.1464 ± 0.0021 | 10986002 | EGC-M (No Edge Features) | 2021-04-03 |
| Semi-Supervised Classification with Graph Convolutional Networks | ✓ Link | 0.1595 ± 0.0018 | No | 0.1461 ± 0.0013 | 12484310 | GCN+virtual node | 2016-09-09 |
| How Powerful are Graph Neural Networks? | ✓ Link | 0.1581 ± 0.0026 | No | 0.1439 ± 0.0020 | 13841815 | GIN+virtual node | 2018-10-01 |
| Do We Need Anisotropic Graph Neural Networks? | ✓ Link | 0.1570 ± 0.0032 | No | 0.1453 ± 0.0025 | 10992050 | PNA (No Edge Features) | 2021-04-03 |
| Graph Attention Networks | ✓ Link | 0.1569 ± 0.0010 | No | 0.1442 ± 0.0017 | 11030210 | GAT | 2017-10-30 |
| Do We Need Anisotropic Graph Neural Networks? | ✓ Link | 0.1552 ± 0.0022 | No | 0.1441 ± 0.0016 | 10971506 | MPNN-Max (No Edge Features) | 2021-04-03 |
| Do We Need Anisotropic Graph Neural Networks? | ✓ Link | 0.1528 ± 0.0025 | No | 0.1427 ± 0.0020 | 11156530 | EGC-S (No Edge Features) | 2021-04-03 |
| Semi-Supervised Classification with Graph Convolutional Networks | ✓ Link | 0.1507 ± 0.0018 | No | 0.1399 ± 0.0017 | 11033210 | GCN | 2016-09-09 |
| How Powerful are Graph Neural Networks? | ✓ Link | 0.1495 ± 0.0023 | No | 0.1376 ± 0.0016 | 12390715 | GIN | 2018-10-01 |
| Hierarchical Graph Representation Learning with Differentiable Pooling | ✓ Link | 0.1401 ± 0.0012 | No | 0.1405 ± 0.0012 | 10095826 | DiffPool w/ graphSAGE | 2018-06-22 |