Feature Transportation Improves Graph Neural Networks

Moshe Eliasof, E. Haber, Eran Treister

Abstract

Graph neural networks (GNNs) have shown remarkable success in learning representations for graph-structured data. However, GNNs still face challenges in modeling complex phenomena that involve feature transportation. In this paper, we propose a novel GNN architecture inspired by Advection-Diffusion-Reaction systems, called ADR-GNN. Advection models feature transportation, while diffusion captures the local smoothing of features, and reaction represents the non-linear transformation between feature channels. We provide an analysis of the qualitative behavior of ADR-GNN, that shows the benefit of combining advection, diffusion, and reaction. To demonstrate its efficacy, we evaluate ADR-GNN on real-world node classification and spatio-temporal datasets, and show that it improves or offers competitive performance compared to state-of-the-art networks.

Benchmark Results

Tested on 8 datasets 30 baseline values reported too low

Tables show presented results in global context (best 3 papers + GCN + MLP).

Classic

Cora

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	GraphSAGE+CNA code ↗ GNN global	0.9418 ±0.0033	Graph Neural Networks Need Cluster-Normalize-Activate Modules code ↗
2	GCN+CNA code ↗ GNN global	0.9366 ±0.0048	Graph Neural Networks Need Cluster-Normalize-Activate Modules code ↗
3	TransformerConv code ↗ GNN global	0.9350 ±0.0058	Graph Neural Networks Need Cluster-Normalize-Activate Modules code ↗
21	GCN GNN baseline actual #21 vs paper #195 174 places worse	0.9004 best reported	actual #21 vs paper #195 174 places worse paper 0.8577 best at pub: SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels (2017) best today: Understanding Heterophily for Graph Neural Networks (ICML 2024) paper 0.8577 Δ at pub 0.0140 Δ today 0.0427
32	ADR-GNN_S GNN ★ proposed	0.8943	This paper
39	SAGE GNN global	0.8911 ±0.0012	Efficient Text-Attributed Graph Learning through Selective Annotation and Graph Alignment
42	C&S GNN baseline	0.8905	This paper
44	GAT GNN baseline actual #44 vs paper #175 131 places worse	0.8897 ±0.0014 best reported	actual #44 vs paper #175 131 places worse paper 0.8637 best at pub: Learned Low Precision Graph Neural Networks (2020) best today: Adaptive Tokenization: On the Hop-Overpriority Problem in Tokenized Graph Learning Models (2025) paper 0.8637 Δ at pub 0.0253 Δ today 0.0260
45	GCNII GNN baseline actual #45 vs paper #66 21 places worse	0.8893 ±0.0137 best reported	actual #45 vs paper #66 21 places worse paper 0.8849 best at pub: Orthogonal Graph Neural Networks (AAAI 2021) best today: Large Language Model Meets Graph Neural Network in Knowledge Distillation (AAAI 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8849 Δ today 0.0044
51	H2GCN GNN baseline actual #51 vs paper #107 56 places worse	0.8787	actual #51 vs paper #107 56 places worse paper 0.8787 prev best: Block Modeling-Guided Graph Convolutional Neural Networks (AAAI 2021) best today: Adaptive Tokenization: On the Hop-Overpriority Problem in Tokenized Graph Learning Models (2025) Improvement within reported uncertainty; not highlighted. paper 0.8787 Δ at pub 0.0018 Δ today 0.0095
55	JKNet GNN baseline actual #55 vs paper #218 163 places worse	0.8870 best reported	actual #55 vs paper #218 163 places worse paper 0.8525 best at pub: Learned Low Precision Graph Neural Networks (2020) paper 0.8525 Δ at pub 0.0345
60	PDE-GCN GNN baseline	0.8860	This paper
62	GREAD GNN baseline	0.8857	This paper
80	ACMII GNN baseline	0.8825	This paper
82	WRGAT GNN baseline actual #82 vs paper #82	0.8820	actual #82 vs paper #82 paper 0.8820 prev best: From Node Interaction to Hop Interaction: New Effective and Scalable Graph Learning Paradigm (2022) Improvement within reported uncertainty; not highlighted. paper 0.8820 Δ at pub 0.0025
87	GRAND GNN baseline actual #87 vs paper #87	0.8815	actual #87 vs paper #87 paper 0.8815 prev best: Graph-Coupled Oscillator Networks (ICML 2022) paper 0.8815 Δ at pub 0.0455
96	GRAFF GNN baseline	0.8801	This paper
103	GGCN GNN baseline	0.8795	This paper
108	APPNP GNN baseline	0.8787	This paper
113	MixHop GNN baseline	0.8761	This paper
138	NSD GNN baseline	0.8714	This paper
168	DMP GNN baseline	0.8652	This paper
170	FLODE GNN baseline actual #170 vs paper #713 543 places worse	0.8644 ±0.0117 best reported	actual #170 vs paper #713 543 places worse paper 0.7807 best today: Latent-Hysteresis Graph ODEs: Modeling Coupled Topology-Feature Evolution via Continuous Phase Transitions (2026) paper 0.7807 Δ today 0.0837
207	Geom-GCN GNN baseline actual #207 vs paper #214 7 places worse	0.8540 ±0.0040 best reported	actual #207 vs paper #214 7 places worse paper 0.8527 best at pub: Non-Local Graph Neural Networks (2020) best today: The Expressive Power of Graph Neural Networks: A Survey (2023) paper 0.8527 Δ at pub 0.0003 Δ today 0.0013
243	LINKX GNN baseline	0.8464	This paper
724	MLP Trad global	0.7768	Understanding Heterophily for Graph Neural Networks
734	Ord. GNN GNN baseline	0.7731	This paper

CiteSeer

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	CNA code ↗ GNN global	0.9575 ±0.0058	Graph Neural Networks Need Cluster-Normalize-Activate Modules code ↗
2	IS-GIB code ↗ Hyb global	0.9390 ±0.0187	Individual and Structural Graph Information Bottlenecks for Out-of-Distribution Generalization code ↗
3	EERM code ↗ GNN global	0.9112 ±0.0145	Individual and Structural Graph Information Bottlenecks for Out-of-Distribution Generalization code ↗
8	Ord. GNN GNN baseline	0.9015	This paper
12	FLODE GNN baseline	0.8902	This paper
20	APPNP GNN baseline actual #20 vs paper #122 102 places worse	0.8142 ±0.0026 best reported	actual #20 vs paper #122 102 places worse paper 0.7653 best at pub: Feature Expansion for Graph Neural Networks (ICML 2023) paper 0.7653 Δ at pub 0.0489
35	GAT GNN baseline actual #35 vs paper #209 174 places worse	0.8080 ±0.0026 best reported	actual #35 vs paper #209 174 places worse paper 0.7432 best at pub: Feature Expansion for Graph Neural Networks (ICML 2023) paper 0.7432 Δ at pub 0.0648
36	GCNII GNN baseline actual #36 vs paper #108 72 places worse	0.8076 ±0.0030 best reported	actual #36 vs paper #108 72 places worse paper 0.7713 best at pub: Feature Expansion for Graph Neural Networks (ICML 2023) paper 0.7713 Δ at pub 0.0363
44	GCN GNN baseline actual #44 vs paper #257 213 places worse	0.8033 ±0.0088 best reported	actual #44 vs paper #257 213 places worse paper 0.7368 best at pub: Gophormer: Ego-Graph Transformer for Node Classification (2021) best today: NTFormer: A Composite Node Tokenized Graph Transformer for Node Classification (2024) paper 0.7368 Δ at pub 0.0575 Δ today 0.0665
61	GraphSAGE GNN global	0.7923 ±0.0053	Gophormer: Ego-Graph Transformer for Node Classification
62	JKNet GNN baseline actual #62 vs paper #145 83 places worse	0.7920 best reported	actual #62 vs paper #145 83 places worse paper 0.7585 best at pub: Tackling Over-Smoothing for General Graph Convolutional Networks (2020) paper 0.7585 Δ at pub 0.0335
79	PDE-GCN GNN baseline	0.7848	This paper
83	ADR-GNN Hyb ★ proposed	0.7836	This paper
88	Geom-GCN GNN baseline actual #88 vs paper #88	0.7799	actual #88 vs paper #88 paper 0.7799 prev best: GBK-GNN: Gated Bi-Kernel Graph Neural Networks for Modeling Both Homophily and Heterophily (WWW 2021) Improvement within reported uncertainty; not highlighted. paper 0.7799 Δ at pub 0.0171
99	GREAD GNN baseline	0.7760	This paper
104	GRAFF GNN baseline	0.7730	This paper
107	NSD GNN baseline	0.7714	This paper
108	GGCN GNN baseline	0.7714	This paper
109	ACMII GNN baseline	0.7712	This paper
111	H2GCN GNN baseline	0.7711	This paper
117	DMP GNN baseline	0.7687	This paper
119	WRGAT GNN baseline	0.7681	This paper
122	GRAND GNN baseline actual #122 vs paper #122	0.7657	actual #122 vs paper #122 paper 0.7657 prev best: Discovering Data Geometry: Energy-Driven Neural Diffusion (ICLR 2023) paper 0.7657 Δ at pub 0.0247
126	MixHop GNN baseline actual #126 vs paper #126	0.7626	actual #126 vs paper #126 paper 0.7626 prev best: Structural Entropy Guided Graph Hierarchical Pooling (ICML 2022) paper 0.7626 Δ at pub 0.0486
127	C&S GNN baseline	0.7622	This paper
312	LINKX GNN baseline actual #312 vs paper #312	0.7319	actual #312 vs paper #312 paper 0.7319 prev best: Uplifting Message Passing Neural Network with Graph Original Information (2022) paper 0.7319 Δ at pub 0.1953
350	MLP code ↗ Trad global	0.7289 ±0.0010	Gumbel-MPNN: Graph Rewiring with Gumbel-Softmax code ↗

PubMed

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	LGTL GT global	0.9518 ±0.0011	Adaptive Tokenization: On the Hop-Overpriority Problem in Tokenized Graph Learning Models
2	MixHop (LO) code ↗ GNN global	0.9513 ±0.0038	LOGIN: A Large Language Model Consulted Graph Neural Network Training Framework code ↗
3	GraphSAGE (LO) code ↗ GNN global	0.9511 ±0.0036	LOGIN: A Large Language Model Consulted Graph Neural Network Training Framework code ↗
16	H2GCN GNN baseline actual #16 vs paper #87 71 places worse	0.9361 ±0.0013 best reported	actual #16 vs paper #87 71 places worse paper 0.8949 best at pub: Optimization-Induced Graph Implicit Nonlinear Diffusion (ICML 2022) best today: Adaptive Tokenization: On the Hop-Overpriority Problem in Tokenized Graph Learning Models (2025) paper 0.8949 Δ today 0.0412
17	APPNP GNN baseline actual #17 vs paper #90 73 places worse	0.9355 ±0.0060 best reported	actual #17 vs paper #90 73 places worse paper 0.8940 best at pub: Tackling Over-Smoothing for General Graph Convolutional Networks (2020) best today: LOGIN: A Large Language Model Consulted Graph Neural Network Training Framework (2024) paper 0.8940 Δ at pub 0.0110 Δ today 0.0415
18	GCN GNN baseline actual #18 vs paper #156 138 places worse	0.9296 ±0.0015 best reported	actual #18 vs paper #156 138 places worse paper 0.8813 best at pub: Tackling Over-Smoothing for General Graph Convolutional Networks (2020) best today: Adaptive Tokenization: On the Hop-Overpriority Problem in Tokenized Graph Learning Models (2025) paper 0.8813 Δ at pub 0.0157 Δ today 0.0483
20	GAT GNN baseline actual #20 vs paper #180 160 places worse	0.9233 ±0.0018 best reported	actual #20 vs paper #180 160 places worse paper 0.8762 best at pub: Tackling Over-Smoothing for General Graph Convolutional Networks (2020) best today: Adaptive Tokenization: On the Hop-Overpriority Problem in Tokenized Graph Learning Models (2025) paper 0.8762 Δ at pub 0.0208 Δ today 0.0471
24	JKNet GNN baseline actual #24 vs paper #112 88 places worse	0.9160 best reported	actual #24 vs paper #112 88 places worse paper 0.8894 best at pub: Tackling Over-Smoothing for General Graph Convolutional Networks (2020) paper 0.8894 Δ at pub 0.0266
52	ADR-GNN_S GNN ★ proposed	0.9055	This paper
58	GCNII GNN baseline actual #58 vs paper #58	0.9030	actual #58 vs paper #58 paper 0.9030 prev best: Graph Neural Networks Inspired by Classical Iterative Algorithms (ICML 2021) paper 0.9030 Δ at pub 0.1000
62	GREAD GNN baseline	0.9023	This paper
69	Geom-GCN GNN baseline	0.9005	This paper
70	MixHop GNN baseline actual #70 vs paper #247 177 places worse	0.8531	actual #70 vs paper #247 177 places worse paper 0.8531 prev best: Graph Contrastive Learning under Heterophily via Graph Filters (2023) best today: Language is All a Graph Needs (2023) Improvement within reported uncertainty; not highlighted. paper 0.8531 Δ at pub 0.0021 Δ today 0.0473
71	GRAFF GNN baseline	0.9004	This paper
74	PDE-GCN GNN baseline	0.8993	This paper
81	C&S GNN baseline	0.8974	This paper
82	ACMII GNN baseline	0.8971	This paper
88	NSD GNN baseline	0.8949	This paper
96	DMP GNN baseline	0.8927	This paper
98	GGCN GNN baseline	0.8915	This paper
109	GRAND GNN baseline actual #109 vs paper #109	0.8902	actual #109 vs paper #109 paper 0.8902 prev best: Framelet Message Passing (2023) paper 0.8902 Δ at pub 0.0932
116	SAGE GNN global	0.8881 ±0.0015	Efficient Text-Attributed Graph Learning through Selective Annotation and Graph Alignment
129	WRGAT GNN baseline	0.8852	This paper
138	Ord. GNN GNN baseline	0.8837	This paper
162	LINKX GNN baseline actual #162 vs paper #162	0.8786	actual #162 vs paper #162 paper 0.8786 prev best: Uplifting Message Passing Neural Network with Graph Original Information (2022) paper 0.8786 Δ at pub 0.2020
210	MLP code ↗ Trad global	0.8690 ±0.0074	On the Equivalence of Graph Convolution and Mixup code ↗
224	FLODE GNN baseline	0.8644	This paper

Heterophilic Graphs

Texas

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	P^2GNN Hyb global	0.9672 ±0.0147	$P^2$GNN: Two Prototype Sets to boost GNN Performance
2	DoG code ↗ ext data GNN global	0.9660 ±0.0210	Diffusion on Graph: Augmentation of Graph Structure for Node Classification code ↗
3	GPF-plus code ↗ ext data GNN global	0.9583 ±0.0419	ProG: A Graph Prompt Learning Benchmark code ↗
18	ADR-GNN_S GNN ★ proposed	0.9361 ±0.0426	This paper
34	MLP Trad global	0.9230 ±0.0070	Design your own universe: a physics-informed agnostic method for enhancing graph neural networks
39	GGCN GNN baseline actual #39 vs paper #148 109 places worse	0.9213 ±0.0305 best reported	actual #39 vs paper #148 109 places worse paper 0.8486 ±0.0455 best at pub: Understanding convolution on graphs via energies (TMLR 2022) best today: Graph Wave Networks (WWW 2025) paper 0.8486 ±0.0455 Δ today 0.0727
40	H2GCN GNN baseline actual #40 vs paper #148 108 places worse	0.9189 ±0.0393 best reported	actual #40 vs paper #148 108 places worse paper 0.8486 ±0.0723 best at pub: How Curvature Enhance the Adaptation Power of Framelet GCNs (2023) best today: GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers (ICML 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8486 ±0.0723 Δ at pub 0.0104 Δ today 0.0703
53	GraphSAGE code ↗ GNN global	0.9000 ±0.0365	Graph Neural Networks Need Cluster-Normalize-Activate Modules code ↗
61	FAGCN GNN baseline actual #61 vs paper #193 132 places worse	0.8961 ±0.0152 best reported	actual #61 vs paper #193 132 places worse paper 0.8243 ±0.0287 best at pub: Node-oriented Spectral Filtering for Graph Neural Networks (2022) paper 0.8243 ±0.0287 Δ at pub 0.0718
70	GCNII GNN baseline actual #70 vs paper #398 328 places worse	0.8918 ±0.0443 best reported	actual #70 vs paper #398 328 places worse paper 0.6946 ±0.0383 best at pub: Convolutional Neural Networks on Graphs with Chebyshev Approximation, Revisited (NeurIPS 2022) best today: Piecewise Constant Spectral Graph Neural Network (TMLR 2025) paper 0.6946 ±0.0383 Δ at pub 0.1100 Δ today 0.1972
82	GRAFF GNN baseline	0.8838 ±0.0453	This paper
83	ACMII-GCN GNN baseline	0.8838 ±0.0343	This paper
96	G^2 GNN baseline	0.8757 ±0.0386	This paper
115	ACMP-GCN Hyb baseline	0.8620 ±0.0300	This paper
121	NSD GNN baseline	0.8595 ±0.0551	This paper
181	MagNet GNN baseline actual #181 vs paper #181	0.8330 ±0.0610	actual #181 vs paper #181 paper 0.8330 ±0.0610 prev best: PyTorch Geometric Signed Directed: A Software Package on Graph Neural Networks for Signed and Directed Graphs (LoG 2022) Improvement within reported uncertainty; not highlighted. paper 0.8330 ±0.0610 Δ at pub 0.0090
211	GAT GNN baseline actual #211 vs paper #533 322 places worse	0.8162 ±0.0645 best reported	actual #211 vs paper #533 322 places worse paper 0.5838 ±0.0663 best at pub: Restructuring Graphs for Higher Homophily via Adaptive Spectral Clustering (AAAI 2022) best today: Equivariance Everywhere All At Once: A Recipe for Graph Foundation Models (2025) paper 0.5838 ±0.0663 Δ at pub 0.2142 Δ today 0.2324
217	GRAND GNN baseline actual #217 vs paper #308 91 places worse	0.8140 ±0.0350 best reported	actual #217 vs paper #308 91 places worse paper 0.7568 ±0.0725 best at pub: Understanding convolution on graphs via energies (TMLR 2022) best today: When Graph Neural Networks Meet Dynamic Mode Decomposition (ICLR 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7568 ±0.0725 Δ today 0.0572
257	GCN GNN baseline actual #257 vs paper #576 319 places worse	0.7933 ±0.0447 best reported	actual #257 vs paper #576 319 places worse paper 0.5216 ±0.0516 best at pub: Restructuring Graphs for Higher Homophily via Adaptive Spectral Clustering (AAAI 2022) best today: Graph Wave Networks (WWW 2025) paper 0.5216 ±0.0516 Δ at pub 0.2624 Δ today 0.2717
325	LINKX GNN baseline	0.7460 ±0.0837	This paper

Actor

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	DiGGR Hyb global	0.4535 ±0.0353	Disentangled Generative Graph Representation Learning
2	TRIGON code ↗ Hyb global	0.4402 ±0.0028	Dynamic Triangulation-Based Graph Rewiring for Graph Neural Networks code ↗
3	DoG code ↗ ext data GNN global	0.4320 ±0.0120	Diffusion on Graph: Augmentation of Graph Structure for Node Classification code ↗
38	GCNII GNN baseline actual #38 vs paper #290 252 places worse	0.4082 ±0.0179 best reported	actual #38 vs paper #290 252 places worse paper 0.3287 ±0.0130 best at pub: Feature Expansion for Graph Neural Networks (ICML 2023) best today: $P^2$GNN: Two Prototype Sets to boost GNN Performance (2026) paper 0.3287 ±0.0130 Δ at pub 0.0574 Δ today 0.0795
42	MLP Trad global	0.4026 ±0.0108	Heterophily-Aware Graph Attention Network
52	GPRGNN GNN baseline actual #52 vs paper #247 195 places worse	0.3992 ±0.0067 best reported	actual #52 vs paper #247 195 places worse paper 0.3463 ±0.0122 best at pub: LON-GNN: Spectral GNNs with Learnable Orthonormal Basis (2023) best today: GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers (ICML 2024) paper 0.3463 ±0.0122 Δ at pub 0.0529 Δ today 0.0529
56	FAGCN GNN baseline actual #56 vs paper #240 184 places worse	0.3970 best reported	actual #56 vs paper #240 184 places worse paper 0.3487 ±0.0135 best at pub: Beyond Low-frequency Information in Graph Convolutional Networks (AAAI 2021) paper 0.3487 ±0.0135 Δ at pub 0.0483
62	GGCN GNN baseline actual #62 vs paper #99 37 places worse	0.3781 ±0.0156	actual #62 vs paper #99 37 places worse paper 0.3781 ±0.0156 prev best: Understanding convolution on graphs via energies (TMLR 2022) Improvement within reported uncertainty; not highlighted. paper 0.3781 ±0.0156 Δ at pub 0.0027
65	ADR-GNN_S GNN ★ proposed	0.3916 ±0.0113	This paper
68	GraphSAGE code ↗ GNN global	0.3899 ±0.0085	A Quasi-Wasserstein Loss for Learning Graph Neural Networks code ↗
71	H2GCN GNN baseline actual #71 vs paper #198 127 places worse	0.3885 ±0.0117 best reported	actual #71 vs paper #198 127 places worse paper 0.3570 ±0.0100 best at pub: ES-GNN: Generalizing Graph Neural Networks Beyond Homophily with Edge Splitting (2022) best today: $P^2$GNN: Two Prototype Sets to boost GNN Performance (2026) paper 0.3570 ±0.0100 Δ at pub 0.0270 Δ today 0.0315
75	WRGAT GNN baseline actual #75 vs paper #158 83 places worse	0.3860 ±0.0110 best reported	actual #75 vs paper #158 83 places worse paper 0.3653 ±0.0077 best at pub: ES-GNN: Generalizing Graph Neural Networks Beyond Homophily with Edge Splitting (2022) paper 0.3653 ±0.0077 Δ at pub 0.0207
87	Geom-GCN GNN baseline actual #87 vs paper #305 218 places worse	0.3810 best reported	actual #87 vs paper #305 218 places worse paper 0.3163 ±0.0115 best at pub: Beyond Low-frequency Information in Graph Convolutional Networks (AAAI 2021) paper 0.3163 ±0.0115 Δ at pub 0.0647
100	NSD GNN baseline actual #100 vs paper #105 5 places worse	0.3781 ±0.0115 best reported	actual #100 vs paper #105 5 places worse paper 0.3779 ±0.0115 best today: Graph Parsing Networks (ICLR 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.3779 ±0.0115 Δ today 0.0002
118	G^2 GNN baseline	0.3730 ±0.0101	This paper
129	GRAFF GNN baseline	0.3711 ±0.0108	This paper
131	ACMII-GCN GNN baseline	0.3709 ±0.0132	This paper
135	GCN GNN baseline actual #135 vs paper #411 276 places worse	0.3690 best reported	actual #135 vs paper #411 276 places worse paper 0.2686 ±0.0110 best at pub: Beyond Low-frequency Information in Graph Convolutional Networks (AAAI 2021) paper 0.2686 ±0.0110 Δ at pub 0.1004
162	MixHop GNN baseline actual #162 vs paper #297 135 places worse	0.3648 ±0.0120 best reported	actual #162 vs paper #297 135 places worse paper 0.3222 ±0.0234 best at pub: Towards Deep Attention in Graph Neural Networks: Problems and Remedies (ICML 2023) paper 0.3222 ±0.0234 Δ at pub 0.0426
180	LINKX GNN baseline	0.3610 ±0.0155	This paper
182	GAT GNN baseline actual #182 vs paper #385 203 places worse	0.3605 ±0.0035 best reported	actual #182 vs paper #385 203 places worse paper 0.2845 ±0.0089 best at pub: Hierarchical Graph Transformer with Adaptive Node Sampling (NeurIPS 2022) paper 0.2845 ±0.0089 Δ at pub 0.0760
203	GRAND GNN baseline	0.3562 ±0.0101	This paper

Chameleon

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	Hetero-S (GAT 8-layer) code ↗ Hyb global	0.8693	The Heterophilic Snowflake Hypothesis: Training and Empowering GNNs for Heterophilic Graphs code ↗
2	Dir-GNN+CNA code ↗ GNN global	0.8586 ±0.0180	Graph Neural Networks Need Cluster-Normalize-Activate Modules code ↗
3	GAT code ↗ GNN global	0.8560	The Heterophilic Snowflake Hypothesis: Training and Empowering GNNs for Heterophilic Graphs code ↗
12	A + D + R Hyb ★ proposed	0.7991	This paper
52	GCN GNN global	0.7033	Understanding Heterophily for Graph Neural Networks
169	MLP Trad global	0.5429	Understanding Heterophily for Graph Neural Networks

Squirrel

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	IBG-NN GNN global	0.7763 ±0.0179	Efficient Learning on Large Graphs using a Densifying Regularity Lemma
2	Dir-GNN+CNA code ↗ GNN global	0.7747 ±0.0128	Graph Neural Networks Need Cluster-Normalize-Activate Modules code ↗
3	DHGNN GNN global	0.7684 ±0.0161	Directed Homophily-Aware Graph Neural Network
15	ADR-GNN_S GNN ★ proposed	0.7254 ±0.0220	This paper
26	ACMII-GCN GNN baseline	0.6740 ±0.0221	This paper
31	G^2 GNN baseline	0.6426 ±0.0238	This paper
39	H2GCN GNN baseline actual #39 vs paper #202 163 places worse	0.6190 ±0.0140 best reported	actual #39 vs paper #202 163 places worse paper 0.3648 ±0.0186 best at pub: Optimization-Induced Graph Implicit Nonlinear Diffusion (ICML 2022) best today: Efficient Learning on Large Graphs using a Densifying Regularity Lemma (2025) paper 0.3648 ±0.0186 Δ at pub 0.0427 Δ today 0.2542
40	LINKX GNN baseline	0.6181 ±0.0180	This paper
46	GCN GNN baseline actual #46 vs paper #267 221 places worse	0.6067 best reported	actual #46 vs paper #267 221 places worse paper 0.2396 ±0.0201 best at pub: Restructuring Graphs for Higher Homophily via Adaptive Spectral Clustering (AAAI 2022) best today: Understanding Heterophily for Graph Neural Networks (ICML 2024) paper 0.2396 ±0.0201 Δ at pub 0.3174 Δ today 0.3671
53	GRAFF GNN baseline	0.5901 ±0.0131	This paper
65	NSD GNN baseline	0.5634 ±0.0132	This paper
67	GAT GNN baseline actual #67 vs paper #241 174 places worse	0.5630 ±0.0220 best reported	actual #67 vs paper #241 174 places worse paper 0.3003 ±0.0155 best at pub: Restructuring Graphs for Higher Homophily via Adaptive Spectral Clustering (AAAI 2022) paper 0.3003 ±0.0155 Δ at pub 0.2627
69	GGCN GNN baseline actual #69 vs paper #71 2 places worse	0.5567 ±0.0187 best reported	actual #69 vs paper #71 2 places worse paper 0.5517 ±0.0158 best at pub: GPNet: Simplifying Graph Neural Networks via Multi-channel Geometric Polynomials (2022) best today: $P^2$GNN: Two Prototype Sets to boost GNN Performance (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.5517 ±0.0158 Δ today 0.0050
76	GPRGNN GNN baseline actual #76 vs paper #235 159 places worse	0.5435 ±0.0087 best reported	actual #76 vs paper #235 159 places worse paper 0.3161 ±0.0124 best at pub: LON-GNN: Spectral GNNs with Learnable Orthonormal Basis (2023) best today: Directed Homophily-Aware Graph Neural Network (2025) paper 0.3161 ±0.0124 Δ at pub 0.1854 Δ today 0.2274
96	WRGAT GNN baseline	0.4885 ±0.0078	This paper
120	SAGE code ↗ GNN global	0.4460 ±0.0130	Scale-aware Message Passing For Graph Node Classification code ↗
127	MixHop GNN baseline	0.4380 ±0.0148	This paper
139	FAGCN GNN baseline	0.4259 ±0.0069	This paper
164	GRAND GNN baseline	0.4005 ±0.0150	This paper
172	MLP code ↗ Trad global	0.3930 ±0.0079	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
176	GCNII GNN baseline actual #176 vs paper #178 2 places worse	0.3881 ±0.0197 best reported	actual #176 vs paper #178 2 places worse paper 0.3847 ±0.0158 best at pub: GPNet: Simplifying Graph Neural Networks via Multi-channel Geometric Polynomials (2022) best today: $P^2$GNN: Two Prototype Sets to boost GNN Performance (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.3847 ±0.0158 Δ today 0.0034
179	Geom-GCN GNN baseline actual #179 vs paper #179	0.3832 ±0.0092	actual #179 vs paper #179 paper 0.3832 ±0.0092 prev best: Gradient Gating for Deep Multi-Rate Learning on Graphs (ICLR 2022) Improvement within reported uncertainty; not highlighted. paper 0.3832 ±0.0092 Δ at pub 0.0017

LINKX Benchmarks

arxiv-year

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	LargeScaleNet code ↗ Hyb global	0.6582 ±0.0036	Scale-aware Message Passing For Graph Node Classification code ↗
2	S² DirGCN global	0.6495 ±0.0033	Spatio-Spectral Graph Neural Networks
3	FaberNet code ↗ GNN global	0.6462 ±0.0101	Scale-aware Message Passing For Graph Node Classification code ↗
6	G^2 GNN baseline	0.6330 ±0.0184	This paper
8	ADR-GNN_S GNN ★ proposed	0.6117 ±0.0154	This paper
30	L Prop (2 hop) GNN baseline	0.4607 ±0.0015	This paper
31	GCN code ↗ GNN global	0.4602 ±0.0026	Scale-aware Message Passing For Graph Node Classification code ↗
40	C&S (2 hop) GNN baseline	0.4217 ±0.0027	This paper
56	MLP code ↗ Trad global	0.3670 ±0.0021	Scale-aware Message Passing For Graph Node Classification code ↗
59	SGC (2 hop) GNN baseline	0.3227 ±0.0006	This paper

SOTA performance (not significantly different from the best) Proposed method (this paper) Baseline / comparison (this paper) Global context (other papers) Best at pub — reported result worse than the best known when published (best reported in green) Best at pub — reported value improved over the best known when published Best today — OK at publication; today's leaderboard best is elsewhere