GNNBenchmark
PATTERN
Structure
14,000Graphs
117Nodes
4,749Edges
2Classes
3Node Feature Dim
Coverage
153Results
1Variants
26Papers
- ALINC: Active Learning for Inductive Node Classification via Graph Sampling ALINC: Active Learning for Inductive Node Classification via Graph Sampling
- GCCM: Enhancing Generative Graph Prediction via Contrastive Consistency Model GCCM: Enhancing Generative Graph Prediction via Contrastive Consistency Model
- DARTS-GT: Differentiable Architecture Search for Graph Transformers with Quantifiable Instance-Specific Interpretability Analysis DARTS-GT: Differentiable Architecture Search for Graph Transformers with Quantifiable Instance-Specific Interpretability Analysis
- TANGO: Graph Neural Dynamics via Learned Energy and Tangential Flows TANGO: Graph Neural Dynamics via Learned Energy and Tangential Flows
- Plain Transformers Can be Powerful Graph Learners Plain Transformers Can be Powerful Graph Learners
- KDD 2024: Benchmarking Positional Encodings for GNNs and Graph Transformers Benchmarking Positional Encodings for GNNs and Graph Transformers KDD 2024
- Graph in Graph Neural Network Graph in Graph Neural Network
- Automatic Graph Topology-Aware Transformer Automatic Graph Topology-Aware Transformer
- RandAlign: A Parameter-Free Method for Regularizing Graph Convolutional Networks RandAlign: A Parameter-Free Method for Regularizing Graph Convolutional Networks
- TMLR 2024: Contextualized Messages Boost Graph Representations Contextualized Messages Boost Graph Representations TMLR 2024
- KDD 2024: Graph Mamba: Towards Learning on Graphs with State Space Models Graph Mamba: Towards Learning on Graphs with State Space Models KDD 2024
- ICML 2023: Recurrent Distance Filtering for Graph Representation Learning Recurrent Distance Filtering for Graph Representation Learning ICML 2023
- Enhancing Graph Neural Networks with Quantum Computed Encodings Enhancing Graph Neural Networks with Quantum Computed Encodings
- IJCAI 2023: IJCAI--23 Formatting Instructions IJCAI--23 Formatting Instructions IJCAI 2023
- ICLR 2023: Are More Layers Beneficial to Graph Transformers? Are More Layers Beneficial to Graph Transformers? ICLR 2023
- JMLR 2023: GNNBenchmark GNNBenchmark JMLR 2023
- TMLR 2022: Empowering GNNs via Edge-Aware Weisfeiler-Leman Algorithm Empowering GNNs via Edge-Aware Weisfeiler-Leman Algorithm TMLR 2022
- NeurIPS 2022: Recipe for a General, Powerful, Scalable Graph Transformer Recipe for a General, Powerful, Scalable Graph Transformer NeurIPS 2022
- ICML 2022: Structure-Aware Transformer for Graph Representation Learning Structure-Aware Transformer for Graph Representation Learning ICML 2022
- TMLR 2022: Rewiring with Positional Encodings for Graph Neural Networks Rewiring with Positional Encodings for Graph Neural Networks TMLR 2022
- TMLR 2022: How Expressive are Transformers in Spectral Domain for Graphs? How Expressive are Transformers in Spectral Domain for Graphs? TMLR 2022
- SoGCN: Second-Order Graph Convolutional Networks SoGCN: Second-Order Graph Convolutional Networks
- Beyond Low-pass Filtering: Graph Convolutional Networks with Automatic Filtering Beyond Low-pass Filtering: Graph Convolutional Networks with Automatic Filtering
- ICLR 2021: Topological Graph Neural Networks Topological Graph Neural Networks ICLR 2021
- NeurIPS 2020: Factorizable Graph Convolutional Networks Factorizable Graph Convolutional Networks NeurIPS 2020
- Formatting Instructions for ICLR 2021 Conference Submissions Formatting Instructions for ICLR 2021 Conference Submissions
Binary node classification Semi-supervised node classification — detect specific graph patterns (2 classes). Uses the benchmark-provided synthetic or superpixel node/edge features. Evaluated by Accuracy.
Dataset details
Task
Semi-supervised node classification — detect specific graph patterns (2 classes). The primary catalog metric is Accuracy.
Data
Semi-supervised node classification — detect specific graph patterns (2 classes).
Features
The benchmark release supplies fixed graph tensors and task-specific node or edge inputs. Image graphs use superpixel appearance and position, synthetic node tasks use generated signals, TSP uses coordinates, and CSL is intentionally featureless.
Splits and evaluation
The catalog records these protocols or variants: Standard split. Evaluation uses Accuracy (higher is better). Exact masks or folds must come from the cited release.
Quirks and pitfalls
Use the released train/validation/test files rather than resplitting. These tasks were designed to expose specific architectural limitations, and preprocessing such as changing superpixels, adding positional encodings, or balancing edge labels changes the benchmark.
Sources
SOTA Progression
new SOTA
matches SOTA at that time
other proposed result
other baseline result
Leaderboard
| # | Model | Accuracy ↑ | ACC | AULC | Paper |
|---|---|---|---|---|---|
| 1 | DeepGraph (48) | 0.9066
±
0.0006 | — | — | |
| 2 | GRIT (our run) | 0.9041
±
0.0023 | — | — | |
| 3 | GRIT 2-QiQRW | 0.9016
±
0.0045 | — | — | |
| 4 | PPGT | 0.8975
±
0.0003 | — | — | |
| 5 | CKGCN | 0.8866
±
0.0014 | — | — | |
| 6 | GRIT 1-CQRW | 0.8720
±
0.0004 | — | — | |
| 7 | GRIT | 0.8720
±
0.0008 | — | — | |
| 8 | TANGO-GPS | 0.8718
±
0.0006 | — | — | |
| 9 | SparseGRIT | 0.8717
±
0.0004 | — | — | |
| 10 | GMN | 0.8714
±
0.0012 | — | — | |
| 11 | SparseGRIT + RWSE | 0.8691
±
0.0003 | — | — | |
| 12 | UGAS | 0.8689
±
0.0200 | — | — | |
| 13 | GIG-GatedGCN | 0.8687
±
0.0001 | — | — | |
| 14 | GraphGPS + RWSE | 0.8687
±
0.0001 | — | — | |
| 15 | GraphGPS | 0.8687
±
0.0001 | — | — | |
| 16 | S4G | 0.8687
±
0.0002 | — | — | |
| 17 | K-Subtree SAT | 0.8687
±
0.0004 | — | — | |
| 18 | GPS + RandAlign | 0.8686
±
0.0001 | — | — | |
| 19 | GIN-AK+ | 0.8685
±
0.0006 | — | — | |
| 20 | GatedGCN (FC) | 0.8685
±
0.0002 | — | — | |
| 21 | SAT | 0.8685
±
0.0004 | — | — | |
| 22 | K-Subgraph SAT | 0.8685
±
0.0004 | — | — | |
| 23 | GatedGCN+RWSE | 0.8683
±
0.0003 | — | — | |
| 24 | EGT | 0.8682
±
0.0002 | — | — | |
| 25 | Exphormer + RWSE | 0.8682
±
0.0004 | — | — | |
| 26 | GraphGPS + RWDIFF | 0.8682
±
0.0006 | — | — | |
| 27 | SparseGRIT + SignNet | 0.8679
±
0.0006 | — | — | |
| 28 | Exphormer + RWDIFF | 0.8679
±
0.0002 | — | — | |
| 29 | GraphGPS + GCKN | 0.8679
±
0.0004 | — | — | |
| 30 | GCCM | 0.8677
±
0.0004 | — | — | |
| 31 | SAN + RandAlign | 0.8677
±
0.0007 | — | — | |
| 32 | GatedGCN + SignNet | 0.8676
±
0.0003 | — | — | |
| 33 | GatedGCN + RWDIFF | 0.8676
±
0.0004 | — | — | |
| 34 | GRED | 0.8676
±
0.0002 | — | — | |
| 35 | GatedGCN + GCKN | 0.8676
±
0.0005 | — | — | |
| 36 | Ours(100K) | 0.8676
±
0.0003 | — | — | |
| 37 | Exphormer + SignNet | 0.8675
±
0.0009 | — | — | |
| 38 | EGTAS | 0.8674
±
0.0005 | — | — | |
| 39 | GraphGPS + SignNet | 0.8673
±
0.0007 | — | — | |
| 40 | Exphormer + ESLapPE | 0.8673
±
0.0002 | — | — | |
| 41 | GPTrans-Nano | 0.8673
±
0.0008 | — | — | |
| 42 | PNA(100K) | 0.8673
±
0.0005 | — | — | |
| 43 | EXPHORMER + GCN | 0.8673
±
0.0004 | — | — | |
| 44 | Exphormer + PPR | 0.8671
±
0.0005 | — | — | |
| 45 | GIN-3-TOGL-1 | 0.8670
±
0.0010 | — | — | |
| 46 | DARTS-GT | 0.8670
±
0.0000 | — | — | |
| 47 | GraphGPS + ESLapPE | 0.8670
±
0.0006 | — | — | |
| 48 | GPS-PCL | 0.8669
±
0.0007 | — | — | |
| 49 | GINE+RWSE | 0.8669
±
0.0008 | — | — | |
| 50 | GPS | 0.8669
±
0.0006 | — | — |
SOTA performance
set a new SOTA at the time
matched the SOTA at the time