Graph Structure Prompt Learning: A Novel Methodology to Improve Performance of Graph Neural Networks

Zhenhua Huang, Kunhao Li, Shaojie Wang, Zhaohong Jia, Wentao Zhu, Sharad Mehrotra

Abstract

Graph neural networks (GNNs) are widely applied in graph data modeling. However, existing GNNs are often trained in a task-driven manner that fails to fully capture the intrinsic nature of the graph structure, resulting in sub-optimal node and graph representations. To address this limitation, we propose a novel Graph structure Prompt Learning method (GPL) to enhance the training of GNNs, which is inspired by prompt mechanisms in natural language processing. GPL employs task-independent graph structure losses to encourage GNNs to learn intrinsic graph characteristics while simultaneously solving downstream tasks, producing higher-quality node and graph representations. In extensive experiments on eleven real-world datasets, after being trained by GPL, GNNs significantly outperform their original performance on node classification, graph classification, and edge prediction tasks (up to 10.28\

Benchmark Results

Tested on 8 datasets 25 baseline values reported too low

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

Classic

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 ↗
44	GCN GNN global	0.8033 ±0.0088	NTFormer: A Composite Node Tokenized Graph Transformer for Node Classification
350	MLP code ↗ Trad global	0.7289 ±0.0010	Gumbel-MPNN: Graph Rewiring with Gumbel-Softmax code ↗
603	GPPT ext data GNN baseline actual #603 vs paper #603	0.7031 ±0.0039	actual #603 vs paper #603 paper 0.7031 ±0.0039 prev best: ProG: A Graph Prompt Learning Benchmark (NeurIPS 2024) paper 0.7031 ±0.0039 Δ at pub 0.3305

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 ↗
18	GCN GNN baseline actual #18 vs paper #228 210 places worse	0.9296 ±0.0015 best reported	actual #18 vs paper #228 210 places worse paper 0.8605 ±0.0018 best at pub: Node Identifiers: Compact, Discrete Representations for Efficient Graph Learning (ICLR 2024) best today: Adaptive Tokenization: On the Hop-Overpriority Problem in Tokenized Graph Learning Models (2025) paper 0.8605 ±0.0018 Δ at pub 0.0399 Δ today 0.0691
20	GAT GNN baseline actual #20 vs paper #283 263 places worse	0.9233 ±0.0018 best reported	actual #20 vs paper #283 263 places worse paper 0.8387 ±0.0030 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 0.8387 ±0.0030 Δ at pub 0.0583 Δ today 0.0846
108	ASDGN+GPL GNN baseline	0.8903 ±0.0012	This paper
115	ASDGN GNN baseline	0.8884 ±0.0078	This paper
116	SAGE GNN global	0.8881 ±0.0015	Efficient Text-Attributed Graph Learning through Selective Annotation and Graph Alignment
125	GraphSAGE GNN baseline actual #125 vs paper #242 117 places worse	0.8864 ±0.0030 best reported	actual #125 vs paper #242 117 places worse paper 0.8557 ±0.0021 best at pub: Non-backtracking Graph Neural Networks (TMLR 2023) paper 0.8557 ±0.0021 Δ at pub 0.0307
139	GCN+PL GNN baseline	0.8834 ±0.0021	This paper
158	FusedGAT+GPL GNN baseline	0.8806 ±0.0021	This paper
178	LightGCN+GPL GNN baseline	0.8767 ±0.0001	This paper
181	ARMA+GPL GNN baseline	0.8762 ±0.0005	This paper
197	ChebNet+GPL GNN baseline	0.8722 ±0.0003	This paper
199	ChebNet GNN baseline actual #199 vs paper #226 27 places worse	0.8718 ±0.0029 best reported	actual #199 vs paper #226 27 places worse paper 0.8633 ±0.0016 best at pub: Non-backtracking Graph Neural Networks (TMLR 2023) paper 0.8633 ±0.0016 Δ at pub 0.0085
210	MLP code ↗ Trad global	0.8690 ±0.0074	On the Equivalence of Graph Convolution and Mixup code ↗
217	ARMA GNN baseline actual #217 vs paper #217	0.8678 ±0.0021	actual #217 vs paper #217 paper 0.8678 ±0.0021 prev best: Graph Neural Networks With Convolutional ARMA Filters (2019) paper 0.8678 ±0.0021 Δ at pub 0.0788
218	UniMP +GPL GNN baseline	0.8673 ±0.0004	This paper
221	GraphSage+GPL GNN baseline	0.8660 ±0.0028	This paper
238	UniMP GT baseline	0.8568 ±0.0026	This paper
240	GAT+GPL GNN baseline	0.8562 ±0.0035	This paper
249	GPPT ext data GNN baseline	0.8528 ±0.0012	This paper
288	FusedGAT GNN baseline	0.8365 ±0.0013	This paper
316	LightGCN GNN baseline	0.8256 ±0.0013	This paper

Amazon-Photo

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	GraphTARIF code ↗ Hyb global	0.9703 ±0.0019	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
2	GAT GNN baseline actual #2 vs paper #508 506 places worse	0.9669 ±0.0014 best reported	actual #2 vs paper #508 506 places worse paper 0.8507 ±0.0025 best at pub: SignGT: Signed Attention-based Graph Transformer for Graph Representation Learning (2023) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.8507 ±0.0025 Δ at pub 0.0991 Δ today 0.1162
3	GAT+T code ↗ Hyb global	0.9660 ±0.0007	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
7	SAGE code ↗ GNN global	0.9643 ±0.0027	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
8	GCN GNN baseline actual #8 vs paper #146 138 places worse	0.9632 ±0.0008 best reported	actual #8 vs paper #146 138 places worse paper 0.9352 ±0.0062 best at pub: SignGT: Signed Attention-based Graph Transformer for Graph Representation Learning (2023) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.9352 ±0.0062 Δ at pub 0.0124 Δ today 0.0280
38	GAT+GPL GNN baseline	0.9535 ±0.0014	This paper
45	FusedGAT+GPL GNN baseline	0.9527 ±0.0030	This paper
49	UniMP +GPL GNN baseline	0.9513 ±0.0062	This paper
56	GraphSAGE GNN baseline actual #56 vs paper #502 446 places worse	0.9508 ±0.0026 best reported	actual #56 vs paper #502 446 places worse paper 0.8596 ±0.1020 best at pub: A Quasi-Wasserstein Loss for Learning Graph Neural Networks (WWW 2023) Difference within reported uncertainty; not counted as underperformance. paper 0.8596 ±0.1020 Δ at pub 0.0912
65	UniMP GT baseline	0.9490 ±0.0037	This paper
73	GraphSage+GPL GNN baseline	0.9456 ±0.0015	This paper
74	ARMA+GPL GNN baseline	0.9452 ±0.0019	This paper
100	GCN+PL GNN baseline	0.9401 ±0.0014	This paper
119	ChebNet GNN baseline actual #119 vs paper #503 384 places worse	0.9377 ±0.0032 best reported	actual #119 vs paper #503 384 places worse paper 0.8571 ±0.0948 best at pub: LON-GNN: Spectral GNNs with Learnable Orthonormal Basis (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.8571 ±0.0948 Δ at pub 0.0806
126	ASDGN+GPL GNN baseline	0.9369 ±0.0026	This paper
128	ChebNet+GPL GNN baseline	0.9367 ±0.0022	This paper
248	GPPT ext data GNN baseline	0.9276 ±0.0026	This paper
261	FusedGAT GNN baseline	0.9269 ±0.0044	This paper
288	LightGCN+GPL GNN baseline	0.9248 ±0.0002	This paper
323	ASDGN GNN baseline	0.9222 ±0.0069	This paper
369	LightGCN GNN baseline	0.9158 ±0.0005	This paper
390	MLP code ↗ Trad global	0.9128 ±0.0063	Clarify Confused Nodes via Separated Learning code ↗
449	ARMA GNN baseline actual #449 vs paper #449	0.9000 ±0.0495	actual #449 vs paper #449 paper 0.9000 ±0.0495 prev best: Measuring and Relieving the Over-smoothing Problem for Graph Neural Networks from the Topological View (AAAI 2019) Improvement within reported uncertainty; not highlighted. paper 0.9000 ±0.0495 Δ at pub 0.0360

Amazon-Computers

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	GraphTARIF code ↗ Hyb global	0.9461 ±0.0017	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
2	GCN GNN baseline actual #2 vs paper #186 184 places worse	0.9412 ±0.0008 best reported	actual #2 vs paper #186 184 places worse paper 0.8849 ±0.0329 best at pub: Clarify Confused Nodes via Separated Learning (2023) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.8849 ±0.0329 Δ at pub 0.0123 Δ today 0.0563
3	GAT GNN baseline actual #3 vs paper #90 87 places worse	0.9398 ±0.0022 best reported	actual #3 vs paper #90 87 places worse paper 0.9012 ±0.0031 best at pub: SignGT: Signed Attention-based Graph Transformer for Graph Representation Learning (2023) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.9012 ±0.0031 Δ at pub 0.0066 Δ today 0.0386
8	SAGE code ↗ GNN global	0.9351 ±0.0006	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
41	FusedGAT+GPL GNN baseline	0.9161 ±0.0031	This paper
46	GraphSAGE GNN baseline actual #46 vs paper #352 306 places worse	0.9120 ±0.0029 best reported	actual #46 vs paper #352 306 places worse paper 0.8399 ±0.0138 best at pub: Polynormer: Polynomial-Expressive Graph Transformer in Linear Time (ICLR 2024) best today: Can TabPFN Compete with GNNs for Node Classification via Graph Tabularization? (LoG 2025) paper 0.8399 ±0.0138 Δ at pub 0.0721 Δ today 0.0721
51	FusedGAT GNN baseline	0.9105 ±0.0067	This paper
61	GAT+GPL GNN baseline	0.9076 ±0.0017	This paper
66	GCN+PL GNN baseline	0.9067 ±0.0025	This paper
115	UniMP +GPL GNN baseline	0.8984 ±0.0067	This paper
151	UniMP GT baseline	0.8916 ±0.0154	This paper
196	ASDGN+GPL GNN baseline	0.8831 ±0.0010	This paper
215	GPPT ext data GNN baseline	0.8783 ±0.0028	This paper
216	ChebNet GNN baseline actual #216 vs paper #435 219 places worse	0.8782 ±0.0072 best reported	actual #216 vs paper #435 219 places worse paper 0.7909 ±0.0417 best at pub: Node-oriented Spectral Filtering for Graph Neural Networks (2022) paper 0.7909 ±0.0417 Δ at pub 0.0873
275	GraphSage+GPL GNN baseline	0.8635 ±0.0081	This paper
295	ChebNet+GPL GNN baseline	0.8573 ±0.0074	This paper
312	ARMA+GPL GNN baseline	0.8507 ±0.0010	This paper
320	MLP code ↗ Trad global	0.8463	GraphFM: A generalist graph transformer that learns transferable representations across diverse … code ↗
339	LightGCN+GPL GNN baseline	0.8428 ±0.0005	This paper
355	LightGCN GNN baseline	0.8388 ±0.0001	This paper
356	ARMA GNN baseline actual #356 vs paper #356	0.8383 ±0.0058	actual #356 vs paper #356 paper 0.8383 ±0.0058 prev best: Adaptive Propagation Graph Convolutional Network (2020) paper 0.8383 ±0.0058 Δ at pub 0.0308
372	ASDGN GNN baseline	0.8348 ±0.0847	This paper

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
16	SortPool+GPL GNN baseline	0.9600 ±0.0200	This paper
23	GIN+GPL GNN baseline	0.9500 ±0.0447	This paper
39	GIN GNN baseline actual #39 vs paper #652 613 places worse	0.9330 ±0.0290 best reported	actual #39 vs paper #652 613 places worse paper 0.7850 ±0.0229 best at pub: ProtGNN: Towards Self-Explaining Graph Neural Networks (AAAI 2021) paper 0.7850 ±0.0229 Δ at pub 0.1480
63	GCN GNN global	0.9220 ±0.0440	Depth-Adaptive Graph Neural Networks via Learnable Bakry-Émery Curvature
101	TopKPool GNN baseline actual #101 vs paper #759 658 places worse	0.9130 ±0.0070 best reported	actual #101 vs paper #759 658 places worse paper 0.7000 ±0.0387 best at pub: Mind the Label Shift of Augmentation-based Graph OOD Generalization (2023) paper 0.7000 ±0.0387 Δ at pub 0.2130
190	SortPool GNN baseline actual #190 vs paper #190	0.9000 ±0.0224	actual #190 vs paper #190 paper 0.9000 ±0.0224 prev best: CogDL: A Comprehensive Library for Graph Deep Learning (WWW 2021) Improvement within reported uncertainty; not highlighted. paper 0.9000 ±0.0224 Δ at pub 0.0275
235	MEWISPool+GPL GNN baseline	0.8947 ±0.0333	This paper
239	DiffPool GNN baseline actual #239 vs paper #421 182 places worse	0.8940 ±0.0460 best reported	actual #239 vs paper #421 182 places worse paper 0.8600 ±0.0539 best at pub: Total Variation Graph Neural Networks (ICML 2022) best today: Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8600 ±0.0539 Δ at pub 0.0040 Δ today 0.0340
277	DiffPool+GPL GNN baseline	0.8900 ±0.0300	This paper
403	MLP code ↗ Trad global	0.8660 ±0.0495	Breaking the Limits of Message Passing Graph Neural Networks code ↗
456	MEWISPool GNN baseline actual #456 vs paper #614 158 places worse	0.8473 ±0.0473 best reported	actual #456 vs paper #614 158 places worse paper 0.8000 ±0.0316 best at pub: Union Subgraph Neural Networks (AAAI 2023) Difference within reported uncertainty; not counted as underperformance. paper 0.8000 ±0.0316 Δ at pub 0.0473
589	EdgePool GNN baseline actual #589 vs paper #747 158 places worse	0.8101 ±0.0082 best reported	actual #589 vs paper #747 158 places worse paper 0.7250 ±0.0250 best at pub: Embedding Graphs on Grassmann Manifold (2022) paper 0.7250 ±0.0250 Δ at pub 0.0851
601	EdgePool+GPL GNN baseline	0.8050 ±0.0150	This paper
602	TopKPool+GPL GNN baseline	0.8050 ±0.0522	This paper
623	SAGPool GNN baseline actual #623 vs paper #766 143 places worse	0.7986 ±0.0236 best reported	actual #623 vs paper #766 143 places worse paper 0.6800 ±0.0458 best at pub: Embedding Graphs on Grassmann Manifold (2022) paper 0.6800 ±0.0458 Δ at pub 0.1186
636	ASAPool+GPL GNN baseline	0.7950 ±0.0350	This paper
687	GPS+GPL Hyb baseline	0.7650 ±0.0320	This paper
722	ASAPool GNN baseline actual #722 vs paper #781 59 places worse	0.7430 ±0.0770 best reported	actual #722 vs paper #781 59 places worse paper 0.6500 ±0.0447 best at pub: Graph Information Bottleneck for Subgraph Recognition (ICLR 2020) best today: Recognizing Predictive Substructures with Subgraph Information Bottleneck (2021) Difference within reported uncertainty; not counted as underperformance. paper 0.6500 ±0.0447 Δ at pub 0.0930 Δ today 0.0930
742	SAGPool+GPL GNN baseline	0.7300 ±0.0871	This paper
772	GPS Hyb baseline	0.6650 ±0.0808	This paper

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 #176 174 places worse	0.9934 best reported	actual #2 vs paper #176 174 places worse paper 0.8107 ±0.0044 best at pub: 1-WL Expressiveness Is (Almost) All You Need (2022) paper 0.8107 ±0.0044 Δ at pub 0.1827
3	GCN GNN global	0.9727	1-WL Expressiveness Is (Almost) All You Need
5	DiffPool GNN baseline actual #5 vs paper #459 454 places worse	0.9210 ±0.0260 best reported	actual #5 vs paper #459 454 places worse paper 0.7241 ±0.0107 best at pub: Where to Mask: Structure-Guided Masking for Graph Masked Autoencoders (IJCAI 2024) paper 0.7241 ±0.0107 Δ at pub 0.1969
131	GraphSAGE code ↗ GNN global	0.8200 ±0.0100	Weisfeiler and Leman Go Infinite: Spectral and Combinatorial Pre-Colorings code ↗
149	GPS Hyb baseline actual #149 vs paper #542 393 places worse	0.8151 ±0.0172 best reported	actual #149 vs paper #542 393 places worse paper 0.6224 ±0.0766 best today: Graph Persistence goes Spectral (NeurIPS 2025) paper 0.6224 ±0.0766 Δ today 0.1927
175	GIN+GPL GNN baseline	0.8117 ±0.0107	This paper
259	SAGPool GNN baseline actual #259 vs paper #462 203 places worse	0.7910 ±0.0300 best reported	actual #259 vs paper #462 203 places worse paper 0.7180 ±0.0290 best at pub: The expressive power of pooling in Graph Neural Networks (NeurIPS 2023) paper 0.7180 ±0.0290 Δ at pub 0.0730
267	EdgePool GNN baseline actual #267 vs paper #462 195 places worse	0.7893 ±0.0222 best reported	actual #267 vs paper #462 195 places worse paper 0.7170 ±0.0049 best at pub: Graph Parsing Networks (ICLR 2024) paper 0.7170 ±0.0049 Δ at pub 0.0723
355	TopKPool GNN baseline actual #355 vs paper #480 125 places worse	0.7706 ±0.0090 best reported	actual #355 vs paper #480 125 places worse paper 0.7029 ±0.0111 best at pub: Embedding Graphs on Grassmann Manifold (2022) paper 0.7029 ±0.0111 Δ at pub 0.0677
407	SortPool+GPL GNN baseline	0.7538 ±0.0086	This paper
415	SortPool GNN baseline actual #415 vs paper #415	0.7496 ±0.0134	actual #415 vs paper #415 paper 0.7496 ±0.0134 prev best: CogDL: A Comprehensive Library for Graph Deep Learning (WWW 2021) Improvement within reported uncertainty; not highlighted. paper 0.7496 ±0.0134 Δ at pub 0.0097
434	MEWISPool GNN baseline actual #434 vs paper #567 133 places worse	0.7421 ±0.0326 best reported	actual #434 vs paper #567 133 places worse paper 0.5634 ±0.0486 best at pub: Union Subgraph Neural Networks (AAAI 2023) paper 0.5634 ±0.0486 Δ at pub 0.1787
446	EdgePool+GPL GNN baseline	0.7331 ±0.0081	This paper
448	DiffPool+GPL GNN baseline	0.7314 ±0.0131	This paper
449	ASAPool GNN baseline actual #449 vs paper #519 70 places worse	0.7310 ±0.0250 best reported	actual #449 vs paper #519 70 places worse paper 0.6594 ±0.0057 best at pub: The expressive power of pooling in Graph Neural Networks (NeurIPS 2023) paper 0.6594 ±0.0057 Δ at pub 0.0716
458	SAGPool+GPL GNN baseline	0.7260 ±0.0137	This paper
470	TopKPool+GPL GNN baseline	0.7114 ±0.0237	This paper
510	MEWISPool+GPL GNN baseline	0.6743 ±0.0600	This paper
514	ASAPool+GPL GNN baseline	0.6640 ±0.0108	This paper
536	GPS+GPL Hyb baseline	0.6323 ±0.0784	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 ↗
6	DiffPool+GPL GNN baseline	0.8196 ±0.0286	This paper
7	DiffPool GNN baseline actual #7 vs paper #7	0.8161 ±0.0232	actual #7 vs paper #7 paper 0.8161 ±0.0232 prev best: Memory-Based Graph Networks (ICLR 2020) paper 0.8161 ±0.0232 Δ at pub 0.0536
11	GIN+GPL GNN baseline	0.8098 ±0.0208	This paper
32	GIN GNN baseline actual #32 vs paper #144 112 places worse	0.7880 ±0.0410 best reported	actual #32 vs paper #144 112 places worse paper 0.7607 ±0.0307 best at pub: Permutation-Invariant graph partitioning: How graph neural networks capture structural interactions? (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.7607 ±0.0307 Δ at pub 0.0273
43	SAGPool+GPL GNN baseline	0.7821 ±0.0215	This paper
58	TopKPool GNN baseline actual #58 vs paper #349 291 places worse	0.7768 best reported	actual #58 vs paper #349 291 places worse paper 0.7339 ±0.0125 best at pub: Memory-Based Graph Networks (ICLR 2020) paper 0.7339 ±0.0125 Δ at pub 0.0429
67	SAGE GNN global	0.7726 ±0.0228	Semi-Supervised Hierarchical Graph Classification
86	TopKPool+GPL GNN baseline	0.7696 ±0.0078	This paper
97	ASAPool+GPL GNN baseline	0.7670 ±0.0129	This paper
134	SortPool+GPL GNN baseline	0.7625 ±0.0091	This paper
149	MLP Trad global	0.7600 ±0.0320	Simplifying Graph Convolutional Networks with Redundancy-Free Neighbors
152	SortPool GNN baseline actual #152 vs paper #152	0.7598 ±0.0176	actual #152 vs paper #152 paper 0.7598 ±0.0176 prev best: Memory-Based Graph Networks (ICLR 2020) Improvement within reported uncertainty; not highlighted. paper 0.7598 ±0.0176 Δ at pub 0.0044
154	SAGPool GNN baseline actual #154 vs paper #175 21 places worse	0.7595 ±0.0452 best reported	actual #154 vs paper #175 21 places worse paper 0.7571 ±0.0108 best at pub: Embedding Graphs on Grassmann Manifold (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.7571 ±0.0108 Δ at pub 0.0024
184	EdgePool GNN baseline actual #184 vs paper #513 329 places worse	0.7560 ±0.0240 best reported	actual #184 vs paper #513 329 places worse paper 0.5840 ±0.0210 best at pub: Embedding Graphs on Grassmann Manifold (2022) paper 0.5840 ±0.0210 Δ at pub 0.1720
323	EdgePool+GPL GNN baseline	0.7384 ±0.0133	This paper
397	ASAPool GNN baseline actual #397 vs paper #397	0.7232 ±0.0291	actual #397 vs paper #397 paper 0.7232 ±0.0291 prev best: Graph Information Bottleneck for Subgraph Recognition (ICLR 2020) Improvement within reported uncertainty; not highlighted. paper 0.7232 ±0.0291 Δ at pub 0.0022
449	MEWISPool+GPL GNN baseline	0.7060 ±0.0286	This paper
464	MEWISPool GNN baseline actual #464 vs paper #464	0.6940 ±0.0144	actual #464 vs paper #464 paper 0.6940 ±0.0144 prev best: Union Subgraph Neural Networks (AAAI 2023) Improvement within reported uncertainty; not highlighted. paper 0.6940 ±0.0144 Δ at pub 0.0130
471	GPS+GPL Hyb baseline	0.6785 ±0.0715	This paper
486	GPS Hyb baseline	0.6446 ±0.0426	This paper

Heterogeneous Graph Benchmarks

DBLP

Global leaderboard →

#	Model	Micro-F1 ↑	Paper
1	LMSPS GNN global	0.9566 ±0.0020	Long-range Meta-path Search on Large-scale Heterogeneous Graphs
2	PSHGCN code ↗ GNN global	0.9561 ±0.0012	Spectral Heterogeneous Graph Convolutions via Positive Noncommutative Polynomials code ↗
3	RpHGNN code ↗ Hyb global	0.9555 ±0.0029	Efficient Heterogeneous Graph Learning via Random Projection code ↗
4	FusedGAT+GPL GNN baseline	0.9544 ±0.0023	This paper
21	GCN GNN baseline actual #21 vs paper #104 83 places worse	0.9460 ±0.0031 best reported	actual #21 vs paper #104 83 places worse paper 0.8517 ±0.0005 best at pub: Graph Neural Networks with a Distribution of Parametrized Graphs (ICML 2023) paper 0.8517 ±0.0005 Δ at pub 0.0943
23	GAT GNN baseline actual #23 vs paper #104 81 places worse	0.9448 ±0.0022 best reported	actual #23 vs paper #104 81 places worse paper 0.8507 ±0.0047 best at pub: Graph Neural Networks with a Distribution of Parametrized Graphs (ICML 2023) paper 0.8507 ±0.0047 Δ at pub 0.0941
42	FusedGAT GNN baseline	0.9340 ±0.0020	This paper
46	GraphSAGE GNN baseline actual #46 vs paper #100 54 places worse	0.9306 ±0.0047 best reported	actual #46 vs paper #100 54 places worse paper 0.8562 ±0.0026 best at pub: Efficient Heterogeneous Graph Learning via Random Projection (2023) paper 0.8562 ±0.0026 Δ at pub 0.0744
88	GIN code ↗ GNN global	0.8720	Model Stealing Attacks Against Inductive Graph Neural Networks code ↗
91	GCN+PL GNN baseline	0.8680 ±0.0001	This paper
92	GAT+GPL GNN baseline	0.8673 ±0.0023	This paper
94	ChebNet+GPL GNN baseline	0.8633 ±0.0002	This paper
95	ARMA+GPL GNN baseline	0.8625 ±0.0006	This paper
96	UniMP +GPL GNN baseline	0.8622 ±0.0018	This paper
97	GraphSage+GPL GNN baseline	0.8607 ±0.0026	This paper
98	ARMA GNN baseline	0.8581 ±0.0013	This paper
100	ChebNet GNN baseline	0.8575 ±0.0012	This paper
101	LightGCN+GPL GNN baseline	0.8555 ±0.0002	This paper
103	UniMP GT baseline	0.8537 ±0.0021	This paper
110	LightGCN GNN baseline	0.8444 ±0.0004	This paper
126	ASDGN+GPL GNN baseline	0.8121 ±0.0020	This paper
134	ASDGN GNN baseline	0.8039 ±0.0028	This paper
176	MLP code ↗ Trad global	0.6447 ±0.0136	GraphControl: Adding Conditional Control to Universal Graph Pre-trained Models for Graph Domain … code ↗

SOTA performance (not significantly different from the best) 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