Node Classification on Pubmed
97.46AccuracySOG
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| SOGbackbone=LLaMA3-3B2026.02 | 97.46 | — | — | — | — | — | 85.5 | — | — | |
| SOGbackbone=LLaMA2-7B2026.02 | 96.27 | — | — | — | — | — | 89.64 | — | — | |
| RGLM-SimilarizerFramework=GTokenLLMs, base_LLM=Vicuna-7B-v1.5-16K2026.03 | 91.61 | — | — | — | — | — | 91.51 | — | — | |
| ACM-Snowball-3Architecture=ACM-Snowball-32024.12 | 91.44 | — | — | — | — | — | — | — | — | |
| RGLM-DecoderFramework=GTokenLLMs, base_LLM=Vicuna-7B-v1.5-16K2026.03 | 91.15 | — | — | — | — | — | 91.07 | — | — | |
| BES2026.06 | 91.03 | — | — | — | — | — | — | — | — | |
| RGLM-DenoiserFramework=GTokenLLMs, base_LLM=Vicuna-7B-v1.5-16K2026.03 | 90.95 | — | — | — | — | — | 90.9 | — | — | |
| G2T-LimiXEvaluation Protocol=FT2025.08 | 90.94 | — | — | — | — | — | — | — | — | |
| FavardGNNcross-validation splits=202023.02 | 90.9 | — | — | — | — | — | — | — | — | |
| tGNN2022.05 | 90.8 | — | — | — | — | — | — | — | — | |
| LGDFramework=GNNs2026.03 | 90.77 | — | — | — | — | — | 90.65 | — | — | |
| Geom-GCN-g2020.05 | 90.7 | — | — | — | — | — | — | — | — | |
| RicciKGE2025.12 | 90.57 | — | — | — | — | — | — | — | — | |
| GraphSAGEFramework=GNNs2026.03 | 90.54 | — | — | — | — | — | 90.51 | — | — | |
| G2T-TabPFNv2Evaluation Protocol=FT2025.08 | 90.52 | — | — | — | — | — | — | — | — | |
| GEOM-GCN2023.12 | 90.49 | — | — | — | — | — | — | — | — | |
| SANEType=AutoGNN2021.12 | 90.47 | — | — | — | — | — | — | — | — | |
| VecFormerCategory=VQ GT2026.02 | 90.47 | — | — | — | — | — | — | — | — | |
| STAGNN2023.10 | 90.46 | — | — | — | — | — | — | — | — | |
| PolynormerCategory=GT2026.02 | 90.44 | — | — | — | — | — | — | — | — | |
| GraphCLFramework=GNNs2026.03 | 90.39 | — | — | — | — | — | 90.28 | — | — | |
| SGFormerCategory=GT2026.02 | 90.37 | — | — | — | — | — | — | — | — | |
| GNRF2025.09 | 90.37 | — | — | — | — | — | — | — | — | |
| FDGATIIsource=This paper, splits=10 averaged2021.10 | 90.3524 | — | — | — | — | — | — | — | — | |
| GCNIIsource=This paper re-test, splits=10 averaged2021.10 | 90.3499 | — | — | — | — | — | — | — | — | |
| UniMPFramework=GNNs2026.03 | 90.34 | — | — | — | — | — | 90.17 | — | — | |
| LLaGAFramework=GTokenLLMs, base_LLM=Vicuna-7B-v1.5-16K2026.03 | 90.34 | — | — | — | — | — | 90.25 | — | — | |
| SOTA2020.10 | 90.3 | — | — | — | — | — | — | — | — | |
| GCNIIEvaluation protocol=fully-supervised2021.05 | 90.3 | — | — | — | — | — | — | — | — | |
| GCNII*Layers=642021.08 | 90.3 | — | — | — | — | — | — | — | — | |
| GCNII*source=Chen et al. (2020), Number of layers=642021.10 | 90.3 | — | — | — | — | — | — | — | — | |
| OptBasisGNNcross-validation splits=202023.02 | 90.3 | — | — | — | — | — | — | — | — | |
| SimP-GCN2023.12 | 90.25 | — | — | — | — | — | — | — | — | |
| BM-GCN2023.12 | 90.25 | — | — | — | — | — | — | — | — | |
| NodeNet2020.06 | 90.21 | — | — | — | — | — | — | — | — | |
| CNAArchitecture=TransformerConv2024.12 | 90.16 | — | — | — | — | — | — | — | — | |
| GCNII2022.06 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNIItrials=102022.05 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNIILayers=Best across layers2021.02 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNIIsplit=48%/32%/20%, random_splits=102023.06 | 90.15 | — | — | — | — | — | — | — | — | |
| ORDERED GNNsplit=48%/32%/20%, random_splits=102023.06 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNII2024.05 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNII2023.05 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNIINumber of Layers=642025.01 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNIIMethod Category=Heterophily-aware baselines2025.12 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNII2025.11 | 90.15 | — | — | — | — | — | — | — | — | |
| GCNII2025.09 | 90.15 | — | — | — | — | — | — | — | — | |
| RCLBackbone=GraphSAGE2023.10 | 90.14 | — | — | — | — | — | — | — | — | |
| Geom-GCN2020.05 | 90.1 | — | — | — | — | — | — | — | — | |
| SADE-GCN2023.05 | 90.07 | — | — | — | — | — | — | — | — | |
| GEOM-GCNEvaluation protocol=fully-supervised2021.05 | 90.05 | — | — | — | — | — | — | — | — | |
| Geom-GCN-I2021.08 | 90.05 | — | — | — | — | — | — | — | — | |
| Geom-GCN-Isource=Chen et al. (2020)2021.10 | 90.05 | — | — | — | — | — | — | — | — | |
| GRAFFsplit=48%/32%/20%, random_splits=102023.06 | 90.04 | — | — | — | — | — | — | — | — | |
| ACM-GCNtrials=102022.05 | 90 | — | — | — | — | — | — | — | — | |
| ACM-GCN2024.05 | 90 | — | — | — | — | — | — | — | — | |
| ACM-GCN2023.05 | 90 | — | — | — | — | — | — | — | — | |
| ACM-GCN2025.09 | 90 | — | — | — | — | — | — | — | — | |
| AutoscaleBase Model=Plain Linear2020.10 | 89.99 | — | — | — | — | — | — | — | — | |
| MPNN-SAGEBackbone=GraphSAGE2026.06 | 89.99 | — | — | — | — | — | — | — | — | |
| GNRF2025.12 | 89.97 | — | — | — | — | — | — | — | — | |
| HiGCNOrder=42023.09 | 89.95 | — | — | — | — | — | — | — | — | |
| Geom-GCN2022.06 | 89.95 | — | — | — | — | — | — | — | — | |
| Geom-GCNLayers=Best across layers2021.02 | 89.95 | — | — | — | — | — | — | — | — | |
| GEOM-GCNsplit=48%/32%/20%, random_splits=102023.06 | 89.95 | — | — | — | — | — | — | — | — | |
| Geom-GCN2023.05 | 89.95 | — | — | — | — | — | — | — | — | |
| Geom-GCNMethod Category=Heterophily-aware baselines2025.12 | 89.95 | — | — | — | — | — | — | — | — | |
| Geom-GCN2025.11 | 89.95 | — | — | — | — | — | — | — | — | |
| AutoscaleBase Model=Linear2020.10 | 89.93 | — | — | — | — | — | — | — | — | |
| PDE-GCNMLayers=162021.08 | 89.93 | — | — | — | — | — | — | — | — | |
| NodeFormerFramework=GNNs2026.03 | 89.92 | — | — | — | — | — | 89.63 | — | — | |
| HiGCNOrder=22023.09 | 89.89 | — | — | — | — | — | — | — | — | |
| HiGCNOrder=12023.09 | 89.83 | — | — | — | — | — | — | — | — | |
| AxelGNN2025.09 | 89.83 | — | — | — | — | — | — | — | — | |
| ACM-GCN+2023.05 | 89.82 | — | — | — | — | — | — | — | — | |
| Cheb2026.06 | 89.8 | — | — | — | — | — | — | — | — | |
| RCLBackbone=GCN2023.10 | 89.79 | — | — | — | — | — | — | — | — | |
| MPNN-GCNBackbone=GCN2026.06 | 89.76 | — | — | — | — | — | — | — | — | |
| FSGNNEvaluation protocol=fully-supervised, Hop count=82021.05 | 89.75 | — | — | — | — | — | — | — | — | |
| BundleChebyT1SDSheaf Type=Bundle, Polynomial=Chebyshev, Order=T12025.11 | 89.75 | — | — | — | — | — | — | — | — | |
| FDiff-scaleBase Model=Plain Linear2020.10 | 89.74 | — | — | — | — | — | — | — | — | |
| HiGCNOrder=32023.09 | 89.73 | — | — | — | — | — | — | — | — | |
| FSGNNEvaluation protocol=fully-supervised, Hop count=32021.05 | 89.73 | — | — | — | — | — | — | — | — | |
| ACMII-GCN++split=48%/32%/20%, random_splits=102023.06 | 89.71 | — | — | — | — | — | — | — | — | |
| NAGphormer2023.10 | 89.7 | — | — | — | — | — | — | — | — | |
| AutoGELType=AutoGNN2021.12 | 89.68 | — | — | — | — | — | — | — | — | |
| Cont Gen-NSDtype=Continuous2022.02 | 89.67 | — | — | — | — | — | — | — | — | |
| GGCN2023.09 | 89.63 | — | — | — | — | — | — | — | — | |
| GloGNNtrials=102022.05 | 89.62 | — | — | — | — | — | — | — | — | |
| JacobiConv2023.02 | 89.62 | — | — | — | — | — | — | — | — | |
| GloGNN2024.05 | 89.62 | — | — | — | — | — | — | — | — | |
| GloGNN2023.05 | 89.62 | — | — | — | — | — | — | — | — | |
| PDE-GCNDLayers=642021.08 | 89.6 | — | — | — | — | — | — | — | — | |
| H2GCNTraining Dimension=512, Supervised=true2023.10 | 89.59 | — | — | — | — | — | — | — | — | |
| H2GCN2023.12 | 89.59 | — | — | — | — | — | — | — | — | |
| H2GCN2024.05 | 89.59 | — | — | — | — | — | — | — | — | |
| GCNIILayers=642021.08 | 89.57 | — | — | — | — | — | — | — | — | |
| GCNIIsource=Chen et al. (2020), Number of layers=642021.10 | 89.57 | — | — | — | — | — | — | — | — | |
| FDiff-scaleBase Model=Linear2020.10 | 89.51 | — | — | — | — | — | — | — | — | |
| PCNet2025.09 | 89.51 | — | — | — | — | — | — | — | — |