LINKX Benchmarks
snap-patents
Structure
1Graphs
2,923,922Nodes
13,975,788Edges
5Classes
269Node Feature Dim
0.073Homophily
Coverage
39Results
1Variants
8Papers
- Scale-aware Message Passing For Graph Node Classification Scale-aware Message Passing For Graph Node Classification
- ICML 2024: Understanding Heterophily for Graph Neural Networks Understanding Heterophily for Graph Neural Networks ICML 2024
- Graph Transformers for Large Graphs Graph Transformers for Large Graphs
- ICLR 2023: HoloNets: Spectral Convolutions do extend to Directed Graphs HoloNets: Spectral Convolutions do extend to Directed Graphs ICLR 2023
- Edge Directionality Improves Learning on Heterophilic Graphs Edge Directionality Improves Learning on Heterophilic Graphs
- ICML 2022: TIDE: Time Derivative Diffusion for Deep Learning on Graphs TIDE: Time Derivative Diffusion for Deep Learning on Graphs ICML 2022
- ICLR 2022: Gradient Gating for Deep Multi-Rate Learning on Graphs Gradient Gating for Deep Multi-Rate Learning on Graphs ICLR 2022
- Non-Local Graph Neural Networks Non-Local Graph Neural Networks
5-class node classification USPTO patent citation graph node classification (5 time-based classes). Uses the released sparse node features and a low-homophily graph. Feature-only baselines are essential because structure can be weakly predictive. Evaluated by Accuracy.
Dataset details
Task
USPTO patent citation graph node classification (5 time-based classes). The primary catalog metric is Accuracy.
Data
USPTO patent citation graph node classification (5 time-based classes). The cataloged artifact reports 1 graphs, 5 target classes or tasks.
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 | FaberNet | 0.7510
±
0.0003 | |
| 2 | LargeScaleNet | 0.7505
±
0.0005 | |
| 3 | FaberNet | 0.7455
±
0.0011 | |
| 4 | DirGNN | 0.7395
±
0.0005 | |
| 5 | dir-gcn | 0.7132
±
0.0006 | |
| 6 | dir-sage | 0.7026
±
0.0014 | |
| 7 | LargeGT-full | 0.7021
±
0.0012 | |
| 8 | G^2-GraphSAGE | 0.6950
±
0.0039 | |
| 9 | LargeGT-local | 0.6819
±
0.0311 | |
| 10 | FSGNN | 0.6507
±
0.0003 | |
| 11 | GloGNN | 0.6209
±
0.0027 | |
| 12 | LINKX | 0.6195
±
0.0012 | |
| 13 | LINK | 0.6039
±
0.0007 | |
| 14 | NAGphormer-δ | 0.6011
±
0.0005 | |
| 15 | ACM-GCN | 0.5514
±
0.0016 | |
| 16 | MixHop | 0.5216
±
0.0009 | |
| 17 | GCN | 0.5102
±
0.0006 | |
| 18 | NLGCN | 0.5080
±
0.0040 | Non-Local Graph Neural Networks
no code |
| 19 | GOAT-full-δ | 0.5028
±
0.0014 | |
| 20 | GraphSAGE-δ | 0.4843
±
0.0021 | |
| 21 | NLGAT | 0.4830
±
0.0020 | Non-Local Graph Neural Networks
no code |
| 22 | GT-sparse-δ | 0.4781
±
0.0000 | |
| 23 | GAT-δ | 0.4592
±
0.0022 | |
| 24 | GCN | 0.4570
±
0.0000 | Non-Local Graph Neural Networks
no code |
| 25 | GAT | 0.4540
±
0.0040 | Non-Local Graph Neural Networks
no code |
| 26 | GOAT-global-δ | 0.4265
±
0.0007 | |
| 27 | GOAT-local-δ | 0.4095
±
0.0016 | |
| 28 | TIDE-s | 0.4075
±
0.0058 | |
| 29 | TIDE-m | 0.4056
±
0.0170 | |
| 30 | GPR-GNN | 0.4019
±
0.0003 | |
| 31 | GRAND | 0.3889
±
0.0042 | |
| 32 | GAT | 0.3876
±
0.0075 | |
| 33 | GCNII | 0.3788
±
0.0069 | |
| 34 | GCN | 0.3684
±
0.0037 | |
| 35 | GCN | 0.3442 | |
| 36 | APPNP | 0.3219
±
0.0007 | |
| 37 | MLP | 0.3150 | |
| 38 | MLP | 0.3130
±
0.0010 | Non-Local Graph Neural Networks
no code |
| 39 | DiffusionNet | 0.3069
±
0.0001 |
SOTA performance
set a new SOTA at the time
matched the SOTA at the time