Scalable Graph Compressed Convolutions

Junshu Sun, Chen Yang, Shuhui Wang, Qingming Huang

Abstract

Designing effective graph neural networks (GNNs) with message passing has two fundamental challenges, i.e., determining optimal message-passing pathways and designing local aggregators. Previous methods of designing optimal pathways are limited with information loss on the input features. On the other hand, existing local aggregators generally fail to extract multi-scale features and approximate diverse operators under limited parameter scales. In contrast to these methods, Euclidean convolution has been proven as an expressive aggregator, making it a perfect candidate for GNN construction. However, the challenges of generalizing Euclidean convolution to graphs arise from the irregular structure of graphs. To bridge the gap between Euclidean space and graph topology, we propose a differentiable method that applies permutations to calibrate input graphs for Euclidean convolution. The permutations constrain all nodes in a row regardless of their input order and therefore enable the flexible generalization of Euclidean convolution to graphs. Based on the graph calibration, we propose the Compressed Convolution Network (CoCN) for hierarchical graph representation learning. CoCN follows local feature-learning and global parameter-sharing mechanisms of convolution neural networks. The whole model can be trained end-to-end, with compressed convolution applied to learn individual node features and their corresponding structure features. CoCN can further borrow successful practices from Euclidean convolution, including residual connection and inception mechanism. We validate CoCN on both node-level and graph-level benchmarks. CoCN achieves superior performance over competitive GNN baselines. Codes are available at https://github.com/sunjss/CoCN.

Benchmark Results

Tested on 10 datasets 37 baseline values reported too low

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

TU Dortmund

MUTAG

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	MSH-GNN GNN global	0.9910 ±0.0030	Multi-Scale Harmonic Encoding for Feature-Wise Graph Message Passing
2	ID-GNN+BEC Hyb global	0.9830 ±0.0360	Depth-Adaptive Graph Neural Networks via Learnable Bakry-Émery Curvature
3	SupCosine Hyb global	0.9830 ±0.0250	Supervised Contrastive Learning with Structure Inference for Graph Classification
39	GIN GNN baseline actual #39 vs paper #586 547 places worse	0.9330 ±0.0290 best reported	actual #39 vs paper #586 547 places worse paper 0.8139 ±0.0153 best at pub: ProtGNN: Towards Self-Explaining Graph Neural Networks (AAAI 2021) paper 0.8139 ±0.0153 Δ at pub 0.1191
55	PATCHY-SAN GNN baseline actual #55 vs paper #279 224 places worse	0.9260 ±0.0420 best reported	actual #55 vs paper #279 224 places worse paper 0.8895 ±0.0421 best at pub: Graph Homomorphism Convolution (ICML 2020) best today: Graph Self-Supervised Learning with Learnable Structural and Positional Encodings (WWW 2025) Difference within reported uncertainty; not counted as underperformance. paper 0.8895 ±0.0421 Δ at pub 0.0355 Δ today 0.0365
63	GCN GNN baseline actual #63 vs paper #759 696 places worse	0.9220 ±0.0440 best reported	actual #63 vs paper #759 696 places worse paper 0.6950 ±0.0178 best at pub: Breaking the Limits of Message Passing Graph Neural Networks (ICML 2021) best today: Depth-Adaptive Graph Neural Networks via Learnable Bakry-Émery Curvature (KDD 2025) paper 0.6950 ±0.0178 Δ at pub 0.1960 Δ today 0.2270
101	TopKPool GNN baseline actual #101 vs paper #766 665 places worse	0.9130 ±0.0070 best reported	actual #101 vs paper #766 665 places worse paper 0.6761 ±0.0336 best at pub: Mind the Label Shift of Augmentation-based Graph OOD Generalization (2023) paper 0.6761 ±0.0336 Δ at pub 0.2369
158	GraphSAGE GNN baseline actual #158 vs paper #504 346 places worse	0.9040 ±0.0780 best reported	actual #158 vs paper #504 346 places worse paper 0.8360 ±0.0960 best at pub: Asynchronous Neural Networks for Learning in Graphs (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.8360 ±0.0960 Δ at pub 0.0680
202	GMT Trad baseline actual #202 vs paper #511 309 places worse	0.8990 ±0.0420 best reported	actual #202 vs paper #511 309 places worse paper 0.8344 ±0.0133 best at pub: Structural Entropy Guided Graph Hierarchical Pooling (ICML 2022) best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) paper 0.8344 ±0.0133 Δ today 0.0646
208	CoCN GNN ★ proposed	0.8989 ±0.0039	This paper
239	DiffPool GNN baseline actual #239 vs paper #640 401 places worse	0.8940 ±0.0460 best reported	actual #239 vs paper #640 401 places worse paper 0.7922 ±0.0102 best at pub: Total Variation Graph Neural Networks (ICML 2022) best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) paper 0.7922 ±0.0102 Δ at pub 0.0718 Δ today 0.1018
245	SEP GNN baseline actual #245 vs paper #436 191 places worse	0.8940 ±0.0610 best reported	actual #245 vs paper #436 191 places worse paper 0.8556 ±0.0109 best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8556 ±0.0109 Δ today 0.0384
367	ASAP GNN baseline actual #367 vs paper #667 300 places worse	0.8740 ±0.0570 best reported	actual #367 vs paper #667 300 places worse paper 0.7783 ±0.0149 best at pub: Multi-grained Semantics-aware Graph Neural Networks (2020) best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) paper 0.7783 ±0.0149 Δ at pub 0.0234 Δ today 0.0957
391	CoCN vanilla Hyb ★ proposed	0.8708 ±0.0017	This paper
403	MLP code ↗ Trad global	0.8660 ±0.0495	Breaking the Limits of Message Passing Graph Neural Networks code ↗
623	SAGPool GNN baseline actual #623 vs paper #734 111 places worse	0.7986 ±0.0236 best reported	actual #623 vs paper #734 111 places worse paper 0.7367 ±0.0428 best at pub: Embedding Graphs on Grassmann Manifold (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.7367 ±0.0428 Δ at pub 0.0619
718	SAGE code ↗ GNN global	0.7490 ±0.0870	Fast Graph Representation Learning with PyTorch Geometric code ↗

NCI1

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	WL-MLP GT global	0.9954	1-WL Expressiveness Is (Almost) All You Need
2	GIN GNN baseline actual #2 vs paper #223 221 places worse	0.9934 best reported	actual #2 vs paper #223 221 places worse paper 0.8000 ±0.0140 best at pub: 1-WL Expressiveness Is (Almost) All You Need (2022) paper 0.8000 ±0.0140 Δ at pub 0.1934
3	GCN GNN baseline actual #3 vs paper #383 380 places worse	0.9727 best reported	actual #3 vs paper #383 380 places worse paper 0.7629 ±0.0179 best at pub: 1-WL Expressiveness Is (Almost) All You Need (2022) paper 0.7629 ±0.0179 Δ at pub 0.2098
5	DiffPool GNN baseline actual #5 vs paper #542 537 places worse	0.9210 ±0.0260 best reported	actual #5 vs paper #542 537 places worse paper 0.6232 ±0.0190 best at pub: Where to Mask: Structure-Guided Masking for Graph Masked Autoencoders (IJCAI 2024) paper 0.6232 ±0.0190 Δ at pub 0.2978
13	CoCN GNN ★ proposed	0.8583 ±0.0086	This paper
80	CoCN vanilla Hyb ★ proposed	0.8289 ±0.0019	This paper
87	PG-GNN GNN baseline	0.8280 ±0.0130	This paper
115	PathNN GNN baseline	0.8230 ±0.0190	This paper
131	GraphSAGE GNN baseline actual #131 vs paper #391 260 places worse	0.8200 ±0.0100 best reported	actual #131 vs paper #391 260 places worse paper 0.7600 ±0.0180 best at pub: Weisfeiler and Leman Go Infinite: Spectral and Combinatorial Pre-Colorings (TMLR 2022) paper 0.7600 ±0.0180 Δ at pub 0.0600
226	GMT Trad baseline actual #226 vs paper #381 155 places worse	0.7990 ±0.0040 best reported	actual #226 vs paper #381 155 places worse paper 0.7635 ±0.0262 best at pub: Structural Entropy Guided Graph Hierarchical Pooling (ICML 2022) best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7635 ±0.0262 Δ today 0.0355
254	SEP GNN baseline actual #254 vs paper #254	0.7935 ±0.0033	actual #254 vs paper #254 paper 0.7935 ±0.0033 prev best: Graph Parsing Networks (ICLR 2024) paper 0.7935 ±0.0033 Δ at pub 0.0126
259	SAGPool GNN baseline actual #259 vs paper #509 250 places worse	0.7910 ±0.0300 best reported	actual #259 vs paper #509 250 places worse paper 0.6745 ±0.0111 best at pub: The expressive power of pooling in Graph Neural Networks (NeurIPS 2023) paper 0.6745 ±0.0111 Δ at pub 0.1165
288	PATCHY-SAN GNN baseline	0.7860 ±0.0190	This paper
309	ASAP GNN baseline actual #309 vs paper #465 156 places worse	0.7821 ±0.0175 best reported	actual #309 vs paper #465 156 places worse paper 0.7148 ±0.0042 best at pub: Multi-grained Semantics-aware Graph Neural Networks (2020) paper 0.7148 ±0.0042 Δ at pub 0.0673
355	TopKPool GNN baseline actual #355 vs paper #510 155 places worse	0.7706 ±0.0090 best reported	actual #355 vs paper #510 155 places worse paper 0.6702 ±0.0225 best at pub: Embedding Graphs on Grassmann Manifold (2022) paper 0.6702 ±0.0225 Δ at pub 0.1004
534	PathNet GNN baseline	0.6410 ±0.0230	This paper

PROTEINS

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	MSH-GNN GNN global	0.9410 ±0.0330	Multi-Scale Harmonic Encoding for Feature-Wise Graph Message Passing
2	GRDL GNN global	0.8260 ±0.0120	Graph Classification via Reference Distribution Learning: Theory and Practice
3	GMN code ↗ Hyb global	0.8225	Memory-Based Graph Networks code ↗
7	DiffPool GNN baseline actual #7 vs paper #132 125 places worse	0.8161 ±0.0232 best reported	actual #7 vs paper #132 125 places worse paper 0.7625 ±0.0100 best at pub: Graph Structure Prompt Learning: A Novel Methodology to Improve Performance of Graph Neural Networks (2024) paper 0.7625 ±0.0100 Δ at pub 0.0536
24	CoCN GNN ★ proposed	0.7948 ±0.0051	This paper
29	ASAP GNN baseline actual #29 vs paper #320 291 places worse	0.7892 ±0.0145 best reported	actual #29 vs paper #320 291 places worse paper 0.7392 ±0.0063 best at pub: Multi-grained Semantics-aware Graph Neural Networks (2020) paper 0.7392 ±0.0063 Δ at pub 0.0500
32	GIN GNN baseline actual #32 vs paper #429 397 places worse	0.7880 ±0.0410 best reported	actual #32 vs paper #429 397 places worse paper 0.7146 ±0.0166 best at pub: Permutation-Invariant graph partitioning: How graph neural networks capture structural interactions? (2023) paper 0.7146 ±0.0166 Δ at pub 0.0734
58	TopKPool GNN baseline actual #58 vs paper #453 395 places worse	0.7768 best reported	actual #58 vs paper #453 395 places worse paper 0.7048 ±0.0101 best at pub: Memory-Based Graph Networks (ICLR 2020) paper 0.7048 ±0.0101 Δ at pub 0.0720
67	SAGE GNN global	0.7726 ±0.0228	Semi-Supervised Hierarchical Graph Classification
90	CoCN vanilla Hyb ★ proposed	0.7686 ±0.0013	This paper
92	PG-GNN GNN baseline	0.7680 ±0.0380	This paper
115	SEP GNN baseline	0.7642 ±0.0039	This paper
148	GCN GNN baseline actual #148 vs paper #354 206 places worse	0.7600 ±0.0320 best reported	actual #148 vs paper #354 206 places worse paper 0.7324 ±0.0073 best at pub: CI-GNN: A Granger Causality-Inspired Graph Neural Network for Interpretable Brain Network-Based Psychiatric Diagnosis (2023) best today: Simplifying Graph Convolutional Networks with Redundancy-Free Neighbors (2025) Difference within reported uncertainty; not counted as underperformance. paper 0.7324 ±0.0073 Δ at pub 0.0276 Δ today 0.0276
149	MLP Trad global	0.7600 ±0.0320	Simplifying Graph Convolutional Networks with Redundancy-Free Neighbors
154	SAGPool GNN baseline actual #154 vs paper #429 275 places worse	0.7595 ±0.0452 best reported	actual #154 vs paper #429 275 places worse paper 0.7156 ±0.0149 best at pub: Embedding Graphs on Grassmann Manifold (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.7156 ±0.0149 Δ at pub 0.0439
155	GraphSAGE GNN baseline actual #155 vs paper #363 208 places worse	0.7590 ±0.0320 best reported	actual #155 vs paper #363 208 places worse paper 0.7300 ±0.0450 best at pub: On Structural Expressive Power of Graph Transformers (KDD 2023) best today: Simplifying Graph Convolutional Networks with Redundancy-Free Neighbors (2025) Difference within reported uncertainty; not counted as underperformance. paper 0.7300 ±0.0450 Δ at pub 0.0290 Δ today 0.0290
225	PathNN GNN baseline	0.7520 ±0.0390	This paper
237	PATCHY-SAN GNN baseline actual #237 vs paper #245 8 places worse	0.7510 ±0.0330 best reported	actual #237 vs paper #245 8 places worse paper 0.7500 ±0.0251 best at pub: Scalable Expressiveness through Preprocessed Graph Perturbations (2024) best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7500 ±0.0251 Δ today 0.0010
236	GMT Trad baseline actual #236 vs paper #239 3 places worse	0.7510 ±0.0060 best reported	actual #236 vs paper #239 3 places worse paper 0.7509 ±0.0059 best at pub: Structural Entropy Guided Graph Hierarchical Pooling (ICML 2022) best today: Multi-Scale Harmonic Encoding for Feature-Wise Graph Message Passing (2025) Difference within reported uncertainty; not counted as underperformance. paper 0.7509 ±0.0059 Δ today 0.0001
453	PathNet GNN baseline	0.7050 ±0.0390	This paper

COLLAB

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	WL-MLP GT global	0.9792	1-WL Expressiveness Is (Almost) All You Need
2	MSH-GNN GNN global	0.9640 ±0.0070	Multi-Scale Harmonic Encoding for Feature-Wise Graph Message Passing
3	CoCN GNN ★ proposed	0.8722 ±0.0013	This paper
5	CoCN vanilla Hyb ★ proposed	0.8615 ±0.0010	This paper
23	GCN GNN baseline actual #23 vs paper #101 78 places worse	0.8260 ±0.0220 best reported	actual #23 vs paper #101 78 places worse paper 0.8059 ±0.0027 best at pub: Scalable Expressiveness through Preprocessed Graph Perturbations (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8059 ±0.0027 Δ at pub 0.0201
30	GIN GNN baseline actual #30 vs paper #174 144 places worse	0.8220 ±0.0210 best reported	actual #30 vs paper #174 144 places worse paper 0.7819 ±0.0063 best at pub: Scalable Expressiveness through Preprocessed Graph Perturbations (2024) paper 0.7819 ±0.0063 Δ at pub 0.0401
33	DiffPool GNN baseline actual #33 vs paper #33	0.8213 ±0.0043	actual #33 vs paper #33 paper 0.8213 ±0.0043 prev best: CogDL: A Comprehensive Library for Graph Deep Learning (WWW 2021) paper 0.8213 ±0.0043 Δ at pub 0.0286
39	SAGPool GNN baseline actual #39 vs paper #176 137 places worse	0.8176 ±0.0157 best reported	actual #39 vs paper #176 137 places worse paper 0.7803 ±0.0031 best at pub: Embedding Graphs on Grassmann Manifold (2022) paper 0.7803 ±0.0031 Δ at pub 0.0373
57	TopKPool GNN baseline actual #57 vs paper #191 134 places worse	0.8140 ±0.0094 best reported	actual #57 vs paper #191 134 places worse paper 0.7758 ±0.0085 best at pub: Embedding Graphs on Grassmann Manifold (2022) paper 0.7758 ±0.0085 Δ at pub 0.0382
69	SEP GNN baseline actual #69 vs paper #69	0.8130 ±0.0020 best reported	actual #69 vs paper #69 paper 0.8128 ±0.0015 best today: Multi-Scale Harmonic Encoding for Feature-Wise Graph Message Passing (2025) Difference within reported uncertainty; not counted as underperformance. paper 0.8128 ±0.0015 Δ today 0.0002
88	PG-GNN GNN baseline	0.8090 ±0.0080	This paper
93	GMT Trad baseline	0.8074 ±0.0054	This paper
136	GraphSAGE GNN baseline actual #136 vs paper #267 131 places worse	0.7970 ±0.0170 best reported	actual #136 vs paper #267 131 places worse paper 0.7390 ±0.0170 best at pub: Quaternion Graph Neural Networks (2020) best today: SignGT: Signed Attention-based Graph Transformer for Graph Representation Learning (2023) paper 0.7390 ±0.0170 Δ at pub 0.0580 Δ today 0.0580
137	SAGE code ↗ GNN global	0.7970 ±0.0170	Fast Graph Representation Learning with PyTorch Geometric code ↗
151	MLP Trad global	0.7920 ±0.0240	Simplifying Graph Convolutional Networks with Redundancy-Free Neighbors
166	ASAP GNN baseline	0.7864 ±0.0050	This paper
275	PATCHY-SAN GNN baseline actual #275 vs paper #279 4 places worse	0.7310 ±0.0270 best reported	actual #275 vs paper #279 4 places worse paper 0.7260 ±0.0215 best at pub: Scalable Expressiveness through Preprocessed Graph Perturbations (2024) best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7260 ±0.0215 Δ today 0.0050

Heterophilic Graphs

Cornell

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	P^2GNN Hyb global	0.9541 ±0.0272	$P^2$GNN: Two Prototype Sets to boost GNN Performance
2	ACM-GCN code ↗ GNN global	0.9480 ±0.0380	Heterophily-informed Message Passing code ↗
3	RF-GCN GNN global	0.9429	Simplifying Graph Convolutional Networks with Redundancy-Free Neighbors
15	GGCN GNN baseline actual #15 vs paper #79 64 places worse	0.9230 ±0.0367 best reported	actual #15 vs paper #79 64 places worse paper 0.8568 ±0.0663 best at pub: Understanding convolution on graphs via energies (TMLR 2022) best today: $P^2$GNN: Two Prototype Sets to boost GNN Performance (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.8568 ±0.0663 Δ today 0.0662
21	APPNP GNN baseline actual #21 vs paper #418 397 places worse	0.9181 ±0.0196 best reported	actual #21 vs paper #418 397 places worse paper 0.5676 ±0.0458 best at pub: LON-GNN: Spectral GNNs with Learnable Orthonormal Basis (2023) paper 0.5676 ±0.0458 Δ at pub 0.3505
24	MLP Trad global	0.9136 ±0.0070	Node-oriented Spectral Filtering for Graph Neural Networks
35	GCNII GNN baseline actual #35 vs paper #203 168 places worse	0.8918 ±0.0396 best reported	actual #35 vs paper #203 168 places worse paper 0.7786 ±0.0379 best at pub: Convolutional Neural Networks on Graphs with Chebyshev Approximation, Revisited (NeurIPS 2022) best today: $P^2$GNN: Two Prototype Sets to boost GNN Performance (2026) paper 0.7786 ±0.0379 Δ at pub 0.0640 Δ today 0.1132
38	FAGCN GNN baseline actual #38 vs paper #216 178 places worse	0.8852 ±0.0133 best reported	actual #38 vs paper #216 178 places worse paper 0.7676 ±0.0587 best at pub: Node-oriented Spectral Filtering for Graph Neural Networks (2022) paper 0.7676 ±0.0587 Δ at pub 0.1176
63	ACMII-GCN GNN baseline actual #63 vs paper #73 10 places worse	0.8649 ±0.0673 best reported	actual #63 vs paper #73 10 places worse paper 0.8595 ±0.0564 best at pub: Graph Neural Reaction Diffusion Models Submitted to the editors June 2023. Accepted in March 2024. (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8595 ±0.0564 Δ at pub 0.0054
71	CoCN vanilla Hyb ★ proposed	0.8622 ±0.0049	This paper
85	ChebNet GNN baseline actual #85 vs paper #244 159 places worse	0.8533 ±0.0104 best reported	actual #85 vs paper #244 159 places worse paper 0.7430 ±0.0746 best at pub: Node-oriented Spectral Filtering for Graph Neural Networks (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.7430 ±0.0746 Δ at pub 0.1103
98	AERO-GNN GNN baseline actual #98 vs paper #163 65 places worse	0.8480 ±0.0330 best reported	actual #98 vs paper #163 65 places worse paper 0.8124 ±0.0680 best at pub: Sign is Not a Remedy: Multiset-to-Multiset Message Passing for Learning on Heterophilic Graphs (ICML 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8124 ±0.0680 Δ at pub 0.0356
123	LINKX GNN baseline actual #123 vs paper #203 80 places worse	0.8346 ±0.0061 best reported	actual #123 vs paper #203 80 places worse paper 0.7784 ±0.0581 best at pub: Differentiable Cluster Graph Neural Network (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7784 ±0.0581 Δ at pub 0.0562
124	CoCN GNN ★ proposed	0.8339 ±0.0165	This paper
138	SAGE GNN global	0.8269	Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
157	GraphSAGE GNN baseline actual #157 vs paper #178 21 places worse	0.8196 ±0.0739 best reported	actual #157 vs paper #178 21 places worse paper 0.8008 ±0.0296 best at pub: Heterophily-Aware Graph Attention Network (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.8008 ±0.0296 Δ at pub 0.0188
170	GCN GNN baseline actual #170 vs paper #417 247 places worse	0.8082 ±0.0360 best reported	actual #170 vs paper #417 247 places worse paper 0.5703 ±0.0467 best at pub: Differentiable Cluster Graph Neural Network (2024) best today: Bayesian Neighborhood Adaptation for Graph Neural Networks (TMLR 2025) paper 0.5703 ±0.0467 Δ at pub 0.2312 Δ today 0.2379
183	MixHop GNN baseline actual #183 vs paper #249 66 places worse	0.7950 ±0.0630 best reported	actual #183 vs paper #249 66 places worse paper 0.7351 ±0.0634 best at pub: Label-Wise Graph Convolutional Network for Heterophilic Graphs (2021) Difference within reported uncertainty; not counted as underperformance. paper 0.7351 ±0.0634 Δ at pub 0.0599
200	GPRGCN GNN baseline	0.7811 ±0.0655	This paper
342	Geom-GCN GNN baseline actual #342 vs paper #372 30 places worse	0.6412 best reported	actual #342 vs paper #372 30 places worse paper 0.6054 ±0.0367 best at pub: Steering Graph Neural Networks with Pinning Control (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.6054 ±0.0367 Δ at pub 0.0358

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 #114 76 places worse	0.4082 ±0.0179 best reported	actual #38 vs paper #114 76 places worse paper 0.3744 ±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 0.3744 ±0.0130 Δ at pub 0.0117 Δ today 0.0338
41	APPNP GNN baseline actual #41 vs paper #351 310 places worse	0.4046 ±0.0064 best reported	actual #41 vs paper #351 310 places worse paper 0.3002 ±0.0089 best at pub: A Quasi-Wasserstein Loss for Learning Graph Neural Networks (WWW 2023) paper 0.3002 ±0.0089 Δ at pub 0.1044
42	MLP Trad global	0.4026 ±0.0108	Heterophily-Aware Graph Attention Network
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 0.3482 ±0.0135 best at pub: Beyond Low-frequency Information in Graph Convolutional Networks (AAAI 2021) paper 0.3482 ±0.0135 Δ at pub 0.0488
62	GGCN GNN baseline actual #62 vs paper #113 51 places worse	0.3929 ±0.0048 best reported	actual #62 vs paper #113 51 places worse paper 0.3754 ±0.0156 best at pub: Feature Transportation Improves Graph Neural Networks (AAAI 2023) best today: $P^2$GNN: Two Prototype Sets to boost GNN Performance (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.3754 ±0.0156 Δ at pub 0.0027 Δ today 0.0175
68	GraphSAGE GNN baseline actual #68 vs paper #234 166 places worse	0.3899 ±0.0085 best reported	actual #68 vs paper #234 166 places worse paper 0.3507 ±0.0015 best at pub: A Quasi-Wasserstein Loss for Learning Graph Neural Networks (WWW 2023) paper 0.3507 ±0.0015 Δ at pub 0.0392
87	Geom-GCN GNN baseline actual #87 vs paper #308 221 places worse	0.3810 best reported	actual #87 vs paper #308 221 places worse paper 0.3159 ±0.0115 best at pub: Beyond Low-frequency Information in Graph Convolutional Networks (AAAI 2021) paper 0.3159 ±0.0115 Δ at pub 0.0651
88	ChebNet GNN baseline actual #88 vs paper #264 176 places worse	0.3802 ±0.0023 best reported	actual #88 vs paper #264 176 places worse paper 0.3411 ±0.0109 best at pub: Node-oriented Spectral Filtering for Graph Neural Networks (2022) paper 0.3411 ±0.0109 Δ at pub 0.0391
131	ACMII-GCN GNN baseline actual #131 vs paper #163 32 places worse	0.3709 ±0.0132 best reported	actual #131 vs paper #163 32 places worse paper 0.3643 ±0.0120 best at pub: Feature Transportation Improves Graph Neural Networks (AAAI 2023) best today: Graph Neural Reaction Diffusion Models Submitted to the editors June 2023. Accepted in March 2024. (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.3643 ±0.0120 Δ at pub 0.0066 Δ today 0.0066
135	GCN GNN baseline actual #135 vs paper #342 207 places worse	0.3690 best reported	actual #135 vs paper #342 207 places worse paper 0.3026 ±0.0079 best at pub: Beyond Low-frequency Information in Graph Convolutional Networks (AAAI 2021) paper 0.3026 ±0.0079 Δ at pub 0.0664
155	CoCN GNN ★ proposed	0.3662 ±0.0085	This paper
157	AERO-GNN GNN baseline	0.3657 ±0.0110	This paper
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
170	CoCN vanilla Hyb ★ proposed	0.3635 ±0.0012	This paper
180	LINKX GNN baseline	0.3610 ±0.0155	This paper
225	GPRGCN GNN baseline	0.3516 ±0.0090	This paper
266	GIN code ↗ GNN global	0.3410 ±0.0030	What Can We Learn From MIMO Graph Convolutions? code ↗

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 ↗
17	CoCN GNN ★ proposed	0.7935 ±0.0142	This paper
18	CoCN exp. ★ proposed	0.7935	This paper
19	CoCN vanilla Hyb ★ proposed	0.7917 ±0.0017	This paper
40	AERO-GNN GNN baseline actual #40 vs paper #40	0.7158 ±0.0240	actual #40 vs paper #40 paper 0.7158 ±0.0240 prev best: Breaking the Entanglement of Homophily and Heterophily in Semi-supervised Node Classification (2023) paper 0.7158 ±0.0240 Δ at pub 0.3058
48	GGCN GNN baseline	0.7114 ±0.0184	This paper
52	GCN GNN baseline actual #52 vs paper #138 86 places worse	0.7033 best reported	actual #52 vs paper #138 86 places worse paper 0.5982 ±0.0258 best at pub: Understanding Heterophily for Graph Neural Networks (ICML 2024) paper 0.5982 ±0.0258 Δ at pub 0.1051
63	ACMII-GCN GNN baseline	0.6846 ±0.0170	This paper
64	LINKX GNN baseline	0.6842 ±0.0138	This paper
79	APPNP GNN baseline actual #79 vs paper #252 173 places worse	0.6670 ±0.0270 best reported	actual #79 vs paper #252 173 places worse paper 0.4044 ±0.0202 best at pub: Restructuring Graphs for Higher Homophily via Adaptive Spectral Clustering (AAAI 2022) paper 0.4044 ±0.0202 Δ at pub 0.2626
103	GCNII GNN baseline	0.6386 ±0.0304	This paper
108	SAGE code ↗ GNN global	0.6340 ±0.0300	Scale-aware Message Passing For Graph Node Classification code ↗
113	GPRGCN GNN baseline	0.6259 ±0.0204	This paper
124	Geom-GCN GNN baseline actual #124 vs paper #137 13 places worse	0.6090 best reported	actual #124 vs paper #137 13 places worse paper 0.6000 ±0.0281 best at pub: Non-Local Graph Neural Networks (2020) best today: Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating (AAAI 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.6000 ±0.0281 Δ at pub 0.0090 Δ today 0.0090
129	MixHop GNN baseline	0.6050 ±0.0253	This paper
140	ChebNet GNN baseline actual #140 vs paper #163 23 places worse	0.5928 ±0.0125 best reported	actual #140 vs paper #163 23 places worse paper 0.5524 ±0.0276 best at pub: LON-GNN: Spectral GNNs with Learnable Orthonormal Basis (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.5524 ±0.0276 Δ at pub 0.0404
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
13	CoCN GNN ★ proposed	0.7298 ±0.0178	This paper
14	CoCN vanilla Hyb ★ proposed	0.7295 ±0.0023	This paper
26	ACMII-GCN GNN baseline actual #26 vs paper #84 58 places worse	0.6740 ±0.0221 best reported	actual #26 vs paper #84 58 places worse paper 0.5180 ±0.0150 best at pub: Feature Transportation Improves Graph Neural Networks (AAAI 2023) best today: Graph Neural Reaction Diffusion Models Submitted to the editors June 2023. Accepted in March 2024. (2024) paper 0.5180 ±0.0150 Δ at pub 0.1560 Δ today 0.1560
40	LINKX GNN baseline	0.6181 ±0.0180	This paper
41	AERO-GNN GNN baseline actual #41 vs paper #41	0.6176 ±0.0240	actual #41 vs paper #41 paper 0.6176 ±0.0240 prev best: Breaking the Entanglement of Homophily and Heterophily in Semi-supervised Node Classification (2023) paper 0.6176 ±0.0240 Δ at pub 0.1996
46	GCN GNN global	0.6067	Understanding Heterophily for Graph Neural Networks
68	APPNP GNN baseline actual #68 vs paper #242 174 places worse	0.5590 ±0.0290 best reported	actual #68 vs paper #242 174 places worse paper 0.2920 ±0.0145 best at pub: Restructuring Graphs for Higher Homophily via Adaptive Spectral Clustering (AAAI 2022) paper 0.2920 ±0.0145 Δ at pub 0.2670
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
107	GPRGCN GNN baseline	0.4631 ±0.0246	This paper
125	ChebNet GNN baseline actual #125 vs paper #125	0.4386 ±0.0164	actual #125 vs paper #125 paper 0.4386 ±0.0164 prev best: Optimizing Polynomial Graph Filters: A Novel Adaptive Krylov Subspace Approach (WWW 2024) paper 0.4386 ±0.0164 Δ at pub 0.0305
127	MixHop GNN baseline	0.4380 ±0.0148	This paper
128	GraphSAGE GNN baseline actual #128 vs paper #199 71 places worse	0.4378 ±0.0190 best reported	actual #128 vs paper #199 71 places worse paper 0.3673 ±0.0121 best at pub: Simplifying approach to Node Classification in Graph Neural Networks (2021) best today: UniGAP: A Universal and Adaptive Graph Upsampling Approach to Mitigate Over-Smoothing in Node Classification Tasks (2024) paper 0.3673 ±0.0121 Δ at pub 0.0488 Δ today 0.0705
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 #181 2 places worse	0.3832 ±0.0092 best reported	actual #179 vs paper #181 2 places worse paper 0.3815 ±0.0092 best at pub: Feature Transportation Improves Graph Neural Networks (AAAI 2023) best today: Graph Neural Reaction Diffusion Models Submitted to the editors June 2023. Accepted in March 2024. (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.3815 ±0.0092 Δ at pub 0.0017 Δ today 0.0017

LRGB

PCQM-Contact

Global leaderboard →

#	Model	MRR ↑	Paper
1	FloydNet code ↗ Hyb global	0.6143	FloydNet: A Learning Paradigm for Global Relational Reasoning code ↗
2	Exphormer code ↗ GT global	0.4737 ±0.0024	Benchmarking Positional Encodings for GNNs and Graph Transformers code ↗
3	Exphormer + LapPE code ↗ GT global	0.4737 ±0.0024	Benchmarking Positional Encodings for GNNs and Graph Transformers code ↗
8	GatedGCN GNN baseline actual #8 vs paper #77 69 places worse	0.4670 ±0.0004 best reported	actual #8 vs paper #77 69 places worse paper 0.3218 ±0.0011 best at pub: Learning Long Range Dependencies on Graphs via Random Walks (ICLR 2024) best today: Benchmarking Positional Encodings for GNNs and Graph Transformers (KDD 2024) paper 0.3218 ±0.0011 Δ at pub 0.1452 Δ today 0.1452
15	GINE GNN baseline actual #15 vs paper #78 63 places worse	0.4617 ±0.0005 best reported	actual #15 vs paper #78 63 places worse paper 0.3180 ±0.0027 best at pub: Recipe for a General, Powerful, Scalable Graph Transformer (NeurIPS 2022) best today: Benchmarking Positional Encodings for GNNs and Graph Transformers (KDD 2024) paper 0.3180 ±0.0027 Δ today 0.1437
16	GIN code ↗ GNN global	0.4617 ±0.0005	Learning Long Range Dependencies on Graphs via Random Walks code ↗
28	GCN GNN baseline actual #28 vs paper #76 48 places worse	0.4526 ±0.0006 best reported	actual #28 vs paper #76 48 places worse paper 0.3234 ±0.0006 best at pub: Learning Long Range Dependencies on Graphs via Random Walks (ICLR 2024) best today: Benchmarking Positional Encodings for GNNs and Graph Transformers (KDD 2024) paper 0.3234 ±0.0006 Δ at pub 0.1292 Δ today 0.1292
36	CoCN exp. ★ proposed	0.3713 ±0.0037	This paper
44	Cache-GNN+RWSE GNN baseline	0.3488 ±0.0008	This paper
48	DRew-GCN GNN baseline	0.3444 ±0.0017	This paper
56	Graph Diffuser Trad baseline	0.3388 ±0.0011	This paper
82	GCNII GNN baseline	0.3161 ±0.0004	This paper

Heterophily Benchmark

Amazon-ratings

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	UniGAP global	0.6375 ±0.0110	UniGAP: A Universal and Adaptive Graph Upsampling Approach to Mitigate Over-Smoothing in Node …
2	H2GCN + HalfHop GNN global	0.5961 ±0.0100	UniGAP: A Universal and Adaptive Graph Upsampling Approach to Mitigate Over-Smoothing in Node …
3	GraphSAGE + HalfHop GNN global	0.5897 ±0.0150	UniGAP: A Universal and Adaptive Graph Upsampling Approach to Mitigate Over-Smoothing in Node …
7	H2GCN GNN baseline actual #7 vs paper #196 189 places worse	0.5681 ±0.0090 best reported	actual #7 vs paper #196 189 places worse paper 0.3647 ±0.0023 best at pub: Unleashing the Potential of Fractional Calculus in Graph Neural Networks with FROND (ICLR 2024) best today: UniGAP: A Universal and Adaptive Graph Upsampling Approach to Mitigate Over-Smoothing in Node Classification Tasks (2024) paper 0.3647 ±0.0023 Δ at pub 0.0489 Δ today 0.2034
12	SAGE code ↗ GNN global	0.5569 ±0.0023	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
14	GAT GNN baseline actual #14 vs paper #105 91 places worse	0.5559 ±0.0027 best reported	actual #14 vs paper #105 91 places worse paper 0.4909 ±0.0063 best at pub: Graph Neural Networks for Graphs with Heterophily: A Survey (2022) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.4909 ±0.0063 Δ today 0.0650
24	GCN GNN baseline actual #24 vs paper #111 87 places worse	0.5411 ±0.0034 best reported	actual #24 vs paper #111 87 places worse paper 0.4870 ±0.0063 best at pub: Graph Neural Networks for Graphs with Heterophily: A Survey (2022) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.4870 ±0.0063 Δ today 0.0541
80	Sp. CoCN ★ proposed	0.5074 ±0.0015	This paper
82	SGC GNN baseline	0.5066 ±0.0048	This paper
84	MLP Trad global	0.5058	Understanding Heterophily for Graph Neural Networks
137	FAGCN GNN baseline actual #137 vs paper #156 19 places worse	0.4632 ±0.0250 best reported	actual #137 vs paper #156 19 places worse paper 0.4412 ±0.0030 best at pub: Unleashing the Potential of Fractional Calculus in Graph Neural Networks with FROND (ICLR 2024) best today: Distributed-Order Fractional Graph Operating Network (NeurIPS 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.4412 ±0.0030 Δ at pub 0.0220 Δ today 0.0220
150	GPRGCN GNN baseline	0.4488 ±0.0034	This paper
164	Seg. CoCN ★ proposed	0.4306 ±0.0021	This paper
196	GloGNN GNN baseline actual #196 vs paper #196	0.3728 ±0.0066 best reported	actual #196 vs paper #196 paper 0.3689 ±0.0014 best at pub: Unleashing the Potential of Fractional Calculus in Graph Neural Networks with FROND (ICLR 2024) best today: Distributed-Order Fractional Graph Operating Network (NeurIPS 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.3689 ±0.0014 Δ at pub 0.0039 Δ today 0.0039

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