LINKX Benchmarks
arxiv-year
Structure
1Graphs
169,343Nodes
1,166,243Edges
5Classes
128Node Feature Dim
0.222Homophily
Coverage
69Results
1Variants
14Papers
- WWW 2026: Frequency-Corrupt Based Graph Self-Supervised Learning Frequency-Corrupt Based Graph Self-Supervised Learning WWW 2026
- KDD 2025: Discrepancy-Aware Graph Mask Auto-Encoder Discrepancy-Aware Graph Mask Auto-Encoder KDD 2025
- Efficient Learning on Large Graphs using a Densifying Regularity Lemma Efficient Learning on Large Graphs using a Densifying Regularity Lemma
- Scale-aware Message Passing For Graph Node Classification Scale-aware Message Passing For Graph Node Classification
- Graph Neural Reaction Diffusion Models Submitted to the editors June 2023. Accepted in March 2024. Graph Neural Reaction Diffusion Models Submitted to the editors June 2023. Accepted in March 2024.
- NeurIPS 2024: Spatio-Spectral Graph Neural Networks Spatio-Spectral Graph Neural Networks NeurIPS 2024
- ICML 2024: Understanding Heterophily for Graph Neural Networks Understanding Heterophily for Graph Neural Networks ICML 2024
- NeurIPS 2023: Simple and Asymmetric Graph Contrastive Learning without Augmentations Simple and Asymmetric Graph Contrastive Learning without Augmentations NeurIPS 2023
- ICLR 2023: HoloNets: Spectral Convolutions do extend to Directed Graphs HoloNets: Spectral Convolutions do extend to Directed Graphs ICLR 2023
- AAAI 2023: Feature Transportation Improves Graph Neural Networks Feature Transportation Improves Graph Neural Networks AAAI 2023
- Uplifting Message Passing Neural Network with Graph Original Information Uplifting Message Passing Neural Network with Graph Original Information
- ICLR 2022: Gradient Gating for Deep Multi-Rate Learning on Graphs Gradient Gating for Deep Multi-Rate Learning on Graphs ICLR 2022
- NeurIPS 2022: Convolutional Neural Networks on Graphs with Chebyshev Approximation, Revisited Convolutional Neural Networks on Graphs with Chebyshev Approximation, Revisited NeurIPS 2022
- Label-Wise Graph Convolutional Network for Heterophilic Graphs Label-Wise Graph Convolutional Network for Heterophilic Graphs
5-class node classification OGB-arXiv node classification by publication year (5 classes, heterophilic variant). Uses the released sparse node features and a low-homophily graph. Evaluated by Accuracy.
Dataset details
Task
OGB-arXiv node classification by publication year (5 classes, heterophilic variant). The primary catalog metric is Accuracy.
Data
The LINKX artifact relabels ogbn-arxiv papers into five publication-year bins on the same citation graph.
Features
Each release supplies a sparse node-feature matrix, graph edges, and node labels. Feature meaning varies by source; structure and features are deliberately evaluated without assuming neighboring labels agree.
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
These are transductive single-graph tasks with fixed random splits. Label homophily is low and class balance varies; feature-only and structure-only baselines are essential, and similarly named OGB/source datasets may use different labels or splits.
Sources
SOTA Progression
new SOTA
matches SOTA at that time
other proposed result
other baseline result
Leaderboard
| # | Model | Accuracy ↑ | Paper |
|---|---|---|---|
| 1 | LargeScaleNet | 0.6582
±
0.0036 | |
| 2 | S² DirGCN | 0.6495
±
0.0033 | |
| 3 | FaberNet | 0.6462
±
0.0101 | |
| 4 | DirGCN | 0.6444
±
0.0027 | |
| 5 | DirGNN | 0.6397
±
0.0030 | |
| 6 | G^2 | 0.6330
±
0.0184 | |
| 7 | G^2-GraphSAGE | 0.6330
±
0.0184 | |
| 8 | ADR-GNN_S | 0.6117
±
0.0154 | |
| 9 | MagNet | 0.6029
±
0.0027 | |
| 10 | IBG-NN | 0.6014 | |
| 11 | RDGNN-I | 0.5846
±
0.0059 | |
| 12 | INGNN | 0.5653
±
0.0015 | |
| 13 | LINKX | 0.5600
±
0.0130 | |
| 14 | GCNs | 0.5580
±
0.0020 | |
| 15 | RDGNN-S | 0.5516
±
0.0046 | |
| 16 | GloGNN | 0.5480
±
0.0030 | |
| 17 | LINK | 0.5397
±
0.0018 | |
| 18 | Polynormer | 0.5250
±
0.0077 | |
| 19 | MixHop | 0.5181
±
0.0017 | |
| 20 | FSGNN | 0.5047
±
0.0021 | |
| 21 | AdaGNN | 0.4949
±
0.0016 | |
| 22 | GAT | 0.4937
±
0.0020 | |
| 23 | H2GCN | 0.4910
±
0.0010 | |
| 24 | ChebNetII | 0.4853
±
0.0031 | |
| 25 | ACM-GCN | 0.4841
±
0.0030 | |
| 26 | ACM-GCN | 0.4737
±
0.0059 | |
| 27 | GraphACL | 0.4721
±
0.0039 | |
| 28 | GCNII | 0.4720
±
0.0030 | |
| 29 | ChebNet | 0.4676
±
0.0024 | |
| 30 | BernNet | 0.4634
±
0.0032 | |
| 31 | MixHop | 0.4610
±
0.0050 | |
| 32 | L Prop (2 hop) | 0.4607
±
0.0015 | |
| 33 | GAT | 0.4605
±
0.0051 | |
| 34 | GCN | 0.4600
±
0.0030 | |
| 35 | GPR-GNN | 0.4597
±
0.0026 | |
| 36 | SP-GCL | 0.4580
±
0.0057 | |
| 37 | GPRGNN | 0.4510
±
0.0020 | |
| 38 | GPR-GNN | 0.4507
±
0.0021 | |
| 39 | HGRL | 0.4392
±
0.0052 | |
| 40 | DSSL | 0.4371
±
0.0054 | |
| 41 | DGMAE | 0.4341
±
0.0064 | |
| 42 | BGRL | 0.4302
±
0.0062 | |
| 43 | GCA | 0.4296
±
0.0039 | |
| 44 | GCA | 0.4252
±
0.0025 | |
| 45 | GCN | 0.4244 | |
| 46 | C&S (2 hop) | 0.4217
±
0.0027 | |
| 47 | FC-GSSL | 0.4200
±
0.0011 | |
| 48 | GraphPAE | 0.4185
±
0.0004 | |
| 49 | BGRL | 0.4143
±
0.0004 | |
| 50 | CCA-SSG | 0.4078
±
0.0001 |
SOTA performance
set a new SOTA at the time
matched the SOTA at the time