Heterophily Benchmark
Tolokers
Structure
1Graphs
11,758Nodes
519,000Edges
2Classes
10Node Feature Dim
0.59Homophily
Coverage
237Results
1Variants
31Papers
- Beyond Oversquashing: Understanding Signal Propagation in GNNs Via Observables Beyond Oversquashing: Understanding Signal Propagation in GNNs Via Observables
- On the Safety of Graph Representation Learning On the Safety of Graph Representation Learning
- ICLR 2026: Learning Posterior Predictive Distributions for Node Classification from Synthetic Graph Priors Learning Posterior Predictive Distributions for Node Classification from Synthetic Graph Priors ICLR 2026
- Neighbourhood Transformer: Switchable Attention for Monophily-Aware Graph Learning Neighbourhood Transformer: Switchable Attention for Monophily-Aware Graph Learning
- ATLAS: Adaptive Topology-based Learning at Scale for Homophilic and Heterophilic Graphs ATLAS: Adaptive Topology-based Learning at Scale for Homophilic and Heterophilic Graphs
- Torch Geometric Pool: the Pytorch library for pooling in Graph Neural Networks Torch Geometric Pool: the Pytorch library for pooling in Graph Neural Networks
- NeurIPS 2025: Relieving the Over-Aggregating Effect in Graph Transformers Relieving the Over-Aggregating Effect in Graph Transformers NeurIPS 2025
- Turning Tabular Foundation Models into Graph Foundation Models Turning Tabular Foundation Models into Graph Foundation Models
- Gumbel-MPNN: Graph Rewiring with Gumbel-Softmax Gumbel-MPNN: Graph Rewiring with Gumbel-Softmax
- TANGO: Graph Neural Dynamics via Learned Energy and Tangential Flows TANGO: Graph Neural Dynamics via Learned Energy and Tangential Flows
- KDD 2025: Discrepancy-Aware Graph Mask Auto-Encoder Discrepancy-Aware Graph Mask Auto-Encoder KDD 2025
- NeurIPS 2025: Semi-supervised Graph Anomaly Detection via Robust Homophily Learning Semi-supervised Graph Anomaly Detection via Robust Homophily Learning NeurIPS 2025
- Equivariance Everywhere All At Once: A Recipe for Graph Foundation Models Equivariance Everywhere All At Once: A Recipe for Graph Foundation Models
- Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
- ICML 2025: Toward Data-centric Directed Graph Learning: An Entropy-driven Approach Toward Data-centric Directed Graph Learning: An Entropy-driven Approach ICML 2025
- TMLR 2025: Heterophily-informed Message Passing Heterophily-informed Message Passing TMLR 2025
- Efficient Learning on Large Graphs using a Densifying Regularity Lemma Efficient Learning on Large Graphs using a Densifying Regularity Lemma
- NeurIPS 2024: Even Sparser Graph Transformers Even Sparser Graph Transformers NeurIPS 2024
- NeurIPS 2024: Unitary convolutions for learning on graphs and groups Unitary convolutions for learning on graphs and groups NeurIPS 2024
- TMLR 2024: SCNode: Spatial and Contextual Coordinates for Graph Representation Learning SCNode: Spatial and Contextual Coordinates for Graph Representation Learning TMLR 2024
- Re-evaluating the Advancements of Heterophilic Graph Learning Re-evaluating the Advancements of Heterophilic Graph Learning
- TopoBench: A Framework for Benchmarking Topological Deep Learning TopoBench: A Framework for Benchmarking Topological Deep Learning
- NeurIPS 2024: on Large Graphs using Intersecting Communities on Large Graphs using Intersecting Communities NeurIPS 2024
- Differentiable Cluster Graph Neural Network Differentiable Cluster Graph Neural Network
- Bundle Neural Networks for message diffusion on graphs Bundle Neural Networks for message diffusion on graphs
- TMLR 2024: Contextualized Messages Boost Graph Representations Contextualized Messages Boost Graph Representations TMLR 2024
- ICLR 2024: Polynormer: Polynomial-Expressive Graph Transformer in Linear Time Polynormer: Polynomial-Expressive Graph Transformer in Linear Time ICLR 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
- LoG 2023: Data Augmentation for Supervised Graph Outlier Detection via Latent Diffusion Models Data Augmentation for Supervised Graph Outlier Detection via Latent Diffusion Models LoG 2023
- ICML 2023: Evolving Computation Graphs Evolving Computation Graphs ICML 2023
- Graph Neural Networks for Graphs with Heterophily: A Survey Graph Neural Networks for Graphs with Heterophily: A Survey
Binary node classification Yandex Toloka worker network binary node classification from the heterophilic benchmark. Uses the corrected benchmark features and ten official splits. Low homophily and class imbalance make split and metric choice important. Evaluated by ROC-AUC.
Dataset details
Task
Yandex Toloka worker network binary node classification from the heterophilic benchmark. The primary catalog metric is ROC-AUC.
Data
Yandex Toloka worker network binary node classification from the heterophilic benchmark. The cataloged artifact reports 1 graphs, 2 target classes or tasks.
Features
The benchmark provides one fixed graph, node features, labels, and ten official splits per dataset. Features are dataset-specific and include text-derived, product, synthetic-grid, or platform-account attributes described by the release.
Splits and evaluation
The catalog records these protocols or variants: Standard split. Evaluation uses ROC-AUC (higher is better). Exact masks or folds must come from the cited release.
Quirks and pitfalls
Use the corrected release and its ten masks. Several tasks are imbalanced and use ROC-AUC rather than accuracy; treating missing relationships as negative evidence or tuning on the test masks invalidates comparison.
Sources
SOTA Progression
new SOTA
matches SOTA at that time
other proposed result
other baseline result
Leaderboard
| # | Model | ROC-AUC ↑ | AUROC | Accuracy | Paper |
|---|---|---|---|---|---|
| 1 | Polynormer-r | 0.8591
±
0.0074 | — | — | |
| 2 | DC-GNN | 0.8588
±
0.0081 | — | — | |
| 3 | TANGO-GPS | 0.8566
±
0.0101 | — | — | |
| 4 | Tango -GPS | 0.8566
±
0.0101 | — | — | |
| 5 | TANGO-GatedGCN | 0.8551
±
0.0098 | — | — | |
| 6 | Tango -GatedGCN | 0.8551
±
0.0098 | — | — | |
| 7 | Polynormer + Wideformer | 0.8533
±
0.0023 | — | — | |
| 8 | Polynormer + Entropy Reg | 0.8521
±
0.0033 | — | — | |
| 9 | Lie Unitary GCN | 0.8518
±
0.0043 | — | — | |
| 10 | Lie UniGCN | 0.8518
±
0.0043 | — | — | |
| 11 | GCN (RWSE) | 0.8511
±
0.0077 | — | — | |
| 12 | Schrödinger | 0.8509
±
0.0017 | — | — | |
| 13 | GAT (RWSE) | 0.8502
±
0.0067 | — | — | |
| 14 | Polynormer | 0.8483
±
0.0072 | — | — | |
| 15 | PD-GNN | 0.8483
±
0.0040 | — | — | |
| 16 | Unitary GCN | 0.8483
±
0.0068 | — | — | |
| 17 | UniGCN | 0.8483
±
0.0068 | — | — | |
| 18 | ECG-GCN | 0.8481
±
0.0025 | — | — | |
| 19 | MLP-ECG-GCN | 0.8481
±
0.0025 | — | — | |
| 20 | GPS -GCN+Performer (RWSE) | 0.8481
±
0.0086 | — | — | |
| 21 | BuNN | 0.8478
±
0.0080 | — | — | |
| 22 | MLPBGRL-ECG-GCN | 0.8473
±
0.0023 | — | — | |
| 23 | GPS -GAT+Transformer (LapPE) | 0.8470
±
0.0056 | — | — | |
| 24 | GraphGPS + Wideformer | 0.8467
±
0.0066 | — | — | |
| 25 | GPS -GAT+Performer (LapPE) | 0.8462
±
0.0054 | — | — | |
| 26 | GMN | 0.8452
±
0.0021 | — | — | |
| 27 | Co-GNN(μ,μ) | 0.8445
±
0.0117 | — | — | |
| 28 | GPS GAT+Performer | 0.8438
±
0.0091 | — | — | |
| 29 | GPS_GAT+Performer (RWSE) | 0.8438
±
0.0091 | — | — | |
| 30 | MLP->GNN-ECG-GCN | 0.8435
±
0.0031 | — | — | |
| 31 | BGRL-ECG-GCN | 0.8425
±
0.0022 | — | — | |
| 32 | ECG-GAT-sep | 0.8423
±
0.0025 | — | — | |
| 33 | BGRL-ECG-GAT-sep | 0.8423
±
0.0025 | — | — | |
| 34 | GraphGPS+Entropy Reg | 0.8423
±
0.0051 | — | — | |
| 35 | MLPBGRL-GAT-sep | 0.8411
±
0.0023 | — | — | |
| 36 | ECG-GT-sep | 0.8400
±
0.0024 | — | — | |
| 37 | MLPBGRL-GT-sep | 0.8400
±
0.0024 | — | — | |
| 38 | MLP-ECG-GAT-sep | 0.8391
±
0.0032 | — | — | |
| 39 | GAT-sep | 0.8378
±
0.0043 | — | — | |
| 40 | Exphormer | 0.8377
±
0.0078 | — | — | |
| 41 | Exphormer+GAT | 0.8377
±
0.0078 | — | — | |
| 42 | IBG-NN | 0.8376
±
0.0051 | — | — | |
| 43 | MLP->GNN-ECG-GAT-sep | 0.8375
±
0.0030 | — | — | |
| 44 | ICG-NN | 0.8373
±
0.0078 Accuracy
0.8373
±
0.0078 | — | 0.8373
±
0.0078 | |
| 45 | GraphGPS | 0.8371
±
0.0048 | — | — | |
| 46 | GPS | 0.8371
±
0.0048 | — | — | |
| 47 | GAT | 0.8370
±
0.0047 | — | — | |
| 48 | GPS + Mamba | 0.8370
±
0.0105 | — | — | |
| 49 | MLP-ECG-GT-sep | 0.8365
±
0.0029 | — | — | |
| 50 | GCN | 0.8364
±
0.0067 | — | — |
SOTA performance
set a new SOTA at the time
matched the SOTA at the time