| Graph Neural Reaction Diffusion Models | | 92.72 ± 5.88 | | RDGNN-I | 2024-06-16 |
| CAT: A Causally Graph Attention Network for Trimming Heterophilic Graph | ✓ Link | 88.8±2.1 | | CATv3-sup | 2023-12-14 |
| Improving Graph Neural Networks with Simple Architecture Design | ✓ Link | 87.84±6.19 | | FSGNN (8-hop) | 2021-05-17 |
| GraphRARE: Reinforcement Learning Enhanced Graph Neural Network with Relative Entropy | | 87.84±4.05 | | H2GCN-RARE (λ=1.0) | 2023-12-15 |
| Ordered GNN: Ordering Message Passing to Deal with Heterophily and Over-smoothing | ✓ Link | 87.03±4.73 | | Ordered GNN | 2023-02-03 |
| Diffusion-Jump GNNs: Homophiliation via Learnable Metric Filters | ✓ Link | 87.03±1.62 | | DJ-GNN | 2023-06-29 |
| UniG-Encoder: A Universal Feature Encoder for Graph and Hypergraph Node Classification | ✓ Link | 86.75±6.56 | | UniG-Encoder | 2023-08-03 |
| Neural Sheaf Diffusion: A Topological Perspective on Heterophily and Oversmoothing in GNNs | ✓ Link | 86.49 ± 7.35 | | Diag-NSD | 2022-02-09 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 86.49 ± 6.73 | | ACMII-GCN++ | 2022-10-14 |
| GCNH: A Simple Method For Representation Learning On Heterophilous Graphs | ✓ Link | 86.49±6.98 | | GCNH | 2023-04-21 |
| Sign is Not a Remedy: Multiset-to-Multiset Message Passing for Learning on Heterophilic Graphs | ✓ Link | 86.48 ± 6.1 | | M2M-GNN | 2024-05-31 |
| Self-attention Dual Embedding for Graphs with Heterophily | | 86.21±5.59 | | SADE-GCN | 2023-05-28 |
| Refining Latent Homophilic Structures over Heterophilic Graphs for Robust Graph Convolution Networks | | 85.96±5.1 | | LHS | 2023-12-27 |
| Finding Global Homophily in Graph Neural Networks When Meeting Heterophily | ✓ Link | 85.95±5.10 | | GloGNN++ | 2022-05-15 |
| Sheaf Neural Networks with Connection Laplacians | ✓ Link | 85.95±7.72 | | Conn-NSD | 2022-06-17 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 85.95 ± 5.64 | | ACMII-GCN | 2022-10-14 |
| Two Sides of the Same Coin: Heterophily and Oversmoothing in Graph Convolutional Neural Networks | ✓ Link | 85.68 ± 6.63 | | GGCN | 2021-02-12 |
| Neural Sheaf Diffusion: A Topological Perspective on Heterophily and Oversmoothing in GNNs | ✓ Link | 85.68 ± 6.51 | | Gen-NSD | 2022-02-09 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 85.68 ± 4.84 | | ACM-GCN+ | 2022-10-14 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 85.68 ± 5.8 | | ACM-GCN++ | 2022-10-14 |
| Transfer Entropy in Graph Convolutional Neural Networks | ✓ Link | 85.68 ± 6.63 | | TE-GCNN | 2024-06-08 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 85.41 ± 5.3 | | ACMII-GCN+ | 2022-10-14 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 85.14 ± 6.07 | | ACM-GCN | 2022-10-14 |
| UniGAP: A Universal and Adaptive Graph Upsampling Approach to Mitigate Over-Smoothing in Node Classification Tasks | ✓ Link | 84.96 ± 5.0 | | H2GCN + UniGAP | 2024-07-28 |
| Non-Local Graph Neural Networks | ✓ Link | 84.9 ± 5.7 | | NLMLP | 2020-05-29 |
| Neural Sheaf Diffusion: A Topological Perspective on Heterophily and Oversmoothing in GNNs | ✓ Link | 84.86 ± 4.71 | | O(d)-NSD | 2022-02-09 |
| Universal Deep GNNs: Rethinking Residual Connection in GNNs from a Path Decomposition Perspective for Preventing the Over-smoothing | | 84.32±7.29 | | UDGNN (GCN) | 2022-05-30 |
| Simple Truncated SVD based Model for Node Classification on Heterophilic Graphs | | 84.05±4.67 | | HLP Concat | 2021-06-24 |
| Finding Global Homophily in Graph Neural Networks When Meeting Heterophily | ✓ Link | 83.51±4.26 | | GloGNN | 2022-05-15 |
| Graph Neural Aggregation-diffusion with Metastability | | 83.3±7.0 | | GRADE-GAT | 2024-03-29 |
| Tree Decomposed Graph Neural Network | ✓ Link | 82.92 ± 6.61 (0, 2-6) | | TDGNN-w | 2021-08-25 |
| Breaking the Entanglement of Homophily and Heterophily in Semi-supervised Node Classification | | 82.9±3.0 | | ADPA | 2023-12-07 |
| Make Heterophily Graphs Better Fit GNN: A Graph Rewiring Approach | | 82.88±5.56 | | GraphSAGE+DHGR | 2022-09-17 |
| FDGATII : Fast Dynamic Graph Attention with Initial Residual and Identity Mapping | ✓ Link | 82.4324 | | FDGATII | 2021-10-21 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 82.43 ± 5.44 | | ACM-SGC-1 | 2022-10-14 |
| Revisiting Heterophily For Graph Neural Networks | ✓ Link | 82.43 ± 5.44 | | ACM-SGC-2 | 2022-10-14 |
| CN-Motifs Perceptive Graph Neural Networks | | 82.38 ± 6.13 | | CNMPGNN | 2021-11-15 |
| Breaking the Limit of Graph Neural Networks by Improving the Assortativity of Graphs with Local Mixing Patterns | ✓ Link | 81.62 ± 3.90 | | WRGAT | 2021-06-11 |
| Beyond Homophily with Graph Echo State Networks | | 81.1±6.0 | | Graph ESN | 2022-10-27 |
| Learn from Heterophily: Heterophilous Information-enhanced Graph Neural Network | ✓ Link | 80.00 ± 4.26 | | HiGNN | 2024-03-26 |
| Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs | ✓ Link | 79.46 ± 4.80 | | H2GCN-2 | 2020-06-20 |
| Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs | ✓ Link | 78.11 ± 6.68 | | H2GCN-1 | 2020-06-20 |
| Adaptive Universal Generalized PageRank Graph Neural Network | ✓ Link | 78.11 ± 6.55 | | GPRGCN | 2020-06-14 |
| Simple and Deep Graph Convolutional Networks | ✓ Link | 77.86 ± 3.79 | | GCNII | 2020-07-04 |
| Large Scale Learning on Non-Homophilous Graphs: New Benchmarks and Strong Simple Methods | ✓ Link | 77.84 ± 5.81 | | LINKX | 2021-10-27 |
| Beyond Low-frequency Information in Graph Convolutional Networks | ✓ Link | 76.76 ± 5.87 | | FAGCN | 2021-01-04 |
| DeltaGNN: Graph Neural Network with Information Flow Control | ✓ Link | 75.67±1.91 | | DeltaGNN - control + DC | 2025-01-10 |
| MixHop: Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing | ✓ Link | 73.51 ± 6.34 | | MixHop | 2019-04-30 |
| Transitivity-Preserving Graph Representation Learning for Bridging Local Connectivity and Role-based Similarity | ✓ Link | 70.0 ±4.44 | | UGT | 2023-08-18 |
| DiffWire: Inductive Graph Rewiring via the Lovász Bound | ✓ Link | 69.04 | | CT-Layer | 2022-06-15 |
| Heterophilic Graph Neural Networks Optimization with Causal Message-passing | | 68.23±2.90 | | GREET+CausalMP | 2024-11-21 |
| The Heterophilic Snowflake Hypothesis: Training and Empowering GNNs for Heterophilic Graphs | ✓ Link | 68.18 | | MGNN + Hetero-S (4 layers) | 2024-06-18 |
| Geom-GCN: Geometric Graph Convolutional Networks | ✓ Link | 60.81 | | Geom-GCN-P | 2020-02-13 |
| DiffWire: Inductive Graph Rewiring via the Lovász Bound | ✓ Link | 58.02 | | CT-Layer (PE) | 2022-06-15 |
| Non-Local Graph Neural Networks | ✓ Link | 57.6 ± 5.5 | | NLGCN | 2020-05-29 |
| Geom-GCN: Geometric Graph Convolutional Networks | ✓ Link | 56.76 | | Geom-GCN-I | 2020-02-13 |
| Geom-GCN: Geometric Graph Convolutional Networks | ✓ Link | 55.68 | | Geom-GCN-S | 2020-02-13 |
| Non-Local Graph Neural Networks | ✓ Link | 54.7 ± 7.6 | | NLGAT | 2020-05-29 |
| Understanding over-squashing and bottlenecks on graphs via curvature | ✓ Link | 54.60±0.39 | | SDRF | 2021-11-29 |
| Beyond Homophily: Structure-aware Path Aggregation Graph Neural Network | ✓ Link | | 91.35 | PathNet | 2022-07-20 |