Rethinking Graph Contrastive Learning through Relative Similarity Preservation

Zhiyuan Ning, Pengfei Wang, Ziyue Qiao, Pengyang Wang, Yuanchun Zhou

Abstract

Graph contrastive learning (GCL) has achieved remarkable success by following the computer vision paradigm of preserving absolute similarity between augmented views. However, this approach faces fundamental challenges in graphs due to their discrete, non-Euclidean nature---view generation often breaks semantic validity and similarity verification becomes unreliable. Through analyzing 11 real-world graphs, we discover a universal pattern transcending the homophily-heterophily dichotomy: label consistency systematically diminishes as structural distance increases, manifesting as smooth decay in homophily graphs and oscillatory decay in heterophily graphs. We establish theoretical guarantees for this pattern through random walk theory, proving label distribution convergence and characterizing the mechanisms behind different decay behaviors. This discovery reveals that graphs naturally encode relative similarity patterns, where structurally closer nodes exhibit collectively stronger semantic relationships. Leveraging this insight, we propose RelGCL, a novel GCL framework with complementary pairwise and listwise implementations that preserve these inherent patterns through collective similarity objectives. Extensive experiments demonstrate that our method consistently outperforms 20 existing approaches across both homophily and heterophily graphs, validating the effectiveness of leveraging natural relative similarity over artificial absolute similarity.

Benchmark Results

Tested on 10 datasets 69 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 #573 552 places worse	0.9004 best reported	actual #21 vs paper #573 552 places worse paper 0.8150 best at pub: Understanding Heterophily for Graph Neural Networks (ICML 2024) paper 0.8150 Δ at pub 0.0854
39	SAGE GNN global	0.8911 ±0.0012	Efficient Text-Attributed Graph Learning through Selective Annotation and Graph Alignment
44	GAT GNN baseline actual #44 vs paper #424 380 places worse	0.8897 ±0.0014 best reported	actual #44 vs paper #424 380 places worse paper 0.8300 ±0.0070 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 0.8300 ±0.0070 Δ at pub 0.0590 Δ today 0.0597
98	SUGRL GNN baseline actual #98 vs paper #382 284 places worse	0.8800 ±0.0010 best reported	actual #98 vs paper #382 284 places worse paper 0.8340 ±0.0050 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.8340 ±0.0050 Δ at pub 0.0460
130	BGRL GNN baseline actual #130 vs paper #447 317 places worse	0.8730 ±0.0010 best reported	actual #130 vs paper #447 317 places worse paper 0.8270 ±0.0060 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.8270 ±0.0060 Δ at pub 0.0460
134	GRACE GNN baseline actual #134 vs paper #536 402 places worse	0.8720 ±0.0020 best reported	actual #134 vs paper #536 402 places worse paper 0.8190 ±0.0040 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.8190 ±0.0040 Δ at pub 0.0530
156	MVGRL GNN baseline actual #156 vs paper #365 209 places worse	0.8680 ±0.0050 best reported	actual #156 vs paper #365 209 places worse paper 0.8350 ±0.0040 best at pub: ENGAGE: Explanation Guided Data Augmentation for Graph Representation Learning (2023) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.8350 ±0.0040 Δ at pub 0.0330 Δ today 0.0330
177	DGI GNN baseline actual #177 vs paper #486 309 places worse	0.8630 ±0.0020 best reported	actual #177 vs paper #486 309 places worse paper 0.8230 ±0.0060 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.8230 ±0.0060 Δ at pub 0.0400
248	CCA-SSG GNN baseline actual #248 vs paper #278 30 places worse	0.8460 ±0.0070 best reported	actual #248 vs paper #278 30 places worse paper 0.8420 ±0.0040 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.8420 ±0.0040 Δ at pub 0.0040
254	RelGCL_List ★ proposed	0.8450 ±0.0040	This paper
253	HomoGCL GNN baseline	0.8450 ±0.0050	This paper
264	RelGCL_Pair ★ proposed	0.8440 ±0.0020	This paper
271	COSTA GNN baseline actual #271 vs paper #389 118 places worse	0.8330 ±0.0030	actual #271 vs paper #389 118 places worse paper 0.8330 ±0.0030 prev best: HomoGCL: Rethinking Homophily in Graph Contrastive Learning (KDD 2023) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.8330 ±0.0030 Δ at pub 0.0110 Δ today 0.0100
280	GraphMAE GNN baseline	0.8420 ±0.0040	This paper
286	GraphACL GNN baseline	0.8420 ±0.0030	This paper
351	G-BT GNN baseline actual #351 vs paper #573 222 places worse	0.8360 best reported	actual #351 vs paper #573 222 places worse paper 0.8150 ±0.0040 best at pub: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) paper 0.8150 ±0.0040 Δ at pub 0.0210
406	gCooL GNN baseline actual #406 vs paper #406	0.8320 ±0.0050	actual #406 vs paper #406 paper 0.8320 ±0.0050 prev best: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) Improvement within reported uncertainty; not highlighted. paper 0.8320 ±0.0050 Δ at pub 0.0040
404	AFGRL GNN baseline actual #404 vs paper #404	0.8320 ±0.0040	actual #404 vs paper #404 paper 0.8320 ±0.0040 prev best: HomoGCL: Rethinking Homophily in Graph Contrastive Learning (KDD 2023) paper 0.8320 ±0.0040 Δ at pub 0.0340
409	SP-GCL GNN baseline	0.8320 ±0.0020	This paper
415	GCA GNN baseline actual #415 vs paper #517 102 places worse	0.8310 ±0.0030 best reported	actual #415 vs paper #517 102 places worse paper 0.8210 ±0.0010 best at pub: Graph Contrastive Learning with Implicit Augmentations (2022) paper 0.8210 ±0.0010 Δ at pub 0.0100
419	AutoSSL GNN baseline	0.8310 ±0.0040	This paper
630	VGAE GNN baseline actual #630 vs paper #756 126 places worse	0.8059 ±0.0035 best reported	actual #630 vs paper #756 126 places worse paper 0.7630 ±0.0020 best at pub: Decoupled Self-supervised Learning for Non-Homophilou Graphs (2022) paper 0.7630 ±0.0020 Δ at pub 0.0429
654	GAE GNN baseline actual #654 vs paper #810 156 places worse	0.8011 ±0.0022 best reported	actual #654 vs paper #810 156 places worse paper 0.7150 ±0.0040 best at pub: Graph Masked Autoencoders with Transformers (2022) paper 0.7150 ±0.0040 Δ at pub 0.0861
724	MLP Trad global	0.7768	Understanding Heterophily for Graph Neural Networks

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 ↗
35	GAT GNN baseline actual #35 vs paper #388 353 places worse	0.8080 ±0.0026 best reported	actual #35 vs paper #388 353 places worse paper 0.7250 ±0.0070 best at pub: Feature Expansion for Graph Neural Networks (ICML 2023) paper 0.7250 ±0.0070 Δ at pub 0.0830
44	GCN GNN baseline actual #44 vs paper #603 559 places worse	0.8033 ±0.0088 best reported	actual #44 vs paper #603 559 places worse paper 0.7030 best at pub: NTFormer: A Composite Node Tokenized Graph Transformer for Node Classification (2024) paper 0.7030 Δ at pub 0.1003
61	GraphSAGE GNN global	0.7923 ±0.0053	Gophormer: Ego-Graph Transformer for Node Classification
66	DGI GNN baseline actual #66 vs paper #474 408 places worse	0.7890 ±0.0020 best reported	actual #66 vs paper #474 408 places worse paper 0.7180 ±0.0070 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.7180 ±0.0070 Δ at pub 0.0710
87	MVGRL GNN baseline actual #87 vs paper #292 205 places worse	0.7808 ±0.0006 best reported	actual #87 vs paper #292 205 places worse paper 0.7330 ±0.0050 best at pub: Node Representation Learning in Graph via Node-to-Neighbourhood Mutual Information Maximization (2022) paper 0.7330 ±0.0050 Δ at pub 0.0478
100	SUGRL GNN baseline actual #100 vs paper #326 226 places worse	0.7760 ±0.0040 best reported	actual #100 vs paper #326 226 places worse paper 0.7300 ±0.0040 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.7300 ±0.0040 Δ at pub 0.0460
136	BGRL GNN baseline actual #136 vs paper #550 414 places worse	0.7600 ±0.0020 best reported	actual #136 vs paper #550 414 places worse paper 0.7110 ±0.0080 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.7110 ±0.0080 Δ at pub 0.0490
165	CCA-SSG GNN baseline actual #165 vs paper #316 151 places worse	0.7540 ±0.0100 best reported	actual #165 vs paper #316 151 places worse paper 0.7310 ±0.0030 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.7310 ±0.0030 Δ at pub 0.0230
201	GRACE GNN baseline actual #201 vs paper #538 337 places worse	0.7450 ±0.0010 best reported	actual #201 vs paper #538 337 places worse paper 0.7120 ±0.0050 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.7120 ±0.0050 Δ at pub 0.0330
209	GraphMAE GNN baseline actual #209 vs paper #279 70 places worse	0.7436 best reported	actual #209 vs paper #279 70 places worse paper 0.7340 ±0.0040 best at pub: Text-space Graph Foundation Models: Comprehensive Benchmarks and New Insights (NeurIPS 2024) paper 0.7340 ±0.0040 Δ at pub 0.0096
257	RelGCL_Pair ★ proposed	0.7370 ±0.0040	This paper
258	GraphACL GNN baseline actual #258 vs paper #260 2 places worse	0.7363 ±0.0022 best reported	actual #258 vs paper #260 2 places worse paper 0.7360 ±0.0020 best today: Discrepancy-Aware Graph Mask Auto-Encoder (KDD 2025) Difference within reported uncertainty; not counted as underperformance. paper 0.7360 ±0.0020 Δ today 0.0003
270	RelGCL_List ★ proposed	0.7360 ±0.0050	This paper
347	G-BT GNN baseline actual #347 vs paper #463 116 places worse	0.7290 ±0.0040 best reported	actual #347 vs paper #463 116 places worse paper 0.7190 ±0.0050 best at pub: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) paper 0.7190 ±0.0050 Δ at pub 0.0100
337	COSTA GNN baseline actual #337 vs paper #446 109 places worse	0.7210 ±0.0030	actual #337 vs paper #446 109 places worse paper 0.7210 ±0.0030 prev best: HomoGCL: Rethinking Homophily in Graph Contrastive Learning (KDD 2023) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.7210 ±0.0030 Δ at pub 0.0140 Δ today 0.0080
350	MLP code ↗ Trad global	0.7289 ±0.0010	Gumbel-MPNN: Graph Rewiring with Gumbel-Softmax code ↗
369	gCooL GNN baseline actual #369 vs paper #369	0.7270 ±0.0040	actual #369 vs paper #369 paper 0.7270 ±0.0040 prev best: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) paper 0.7270 ±0.0040 Δ at pub 0.0070
374	AFGRL GNN baseline actual #374 vs paper #374	0.7260 ±0.0030	actual #374 vs paper #374 paper 0.7260 ±0.0030 prev best: HomoGCL: Rethinking Homophily in Graph Contrastive Learning (KDD 2023) paper 0.7260 ±0.0030 Δ at pub 0.0320
427	HomoGCL GNN baseline	0.7230 ±0.0070	This paper
454	AutoSSL GNN baseline	0.7210 ±0.0040	This paper
459	SP-GCL GNN baseline	0.7200 ±0.0040	This paper
505	GCA GNN baseline actual #505 vs paper #528 23 places worse	0.7150 ±0.0030 best reported	actual #505 vs paper #528 23 places worse paper 0.7130 ±0.0040 best at pub: Topology Reorganized Graph Contrastive Learning with Mitigating Semantic Drift (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7130 ±0.0040 Δ at pub 0.0020
625	VGAE GNN baseline actual #625 vs paper #725 100 places worse	0.6990 ±0.0057 best reported	actual #625 vs paper #725 100 places worse paper 0.6680 ±0.0020 best at pub: Decoupled Self-supervised Learning for Non-Homophilou Graphs (2022) paper 0.6680 ±0.0020 Δ at pub 0.0310
739	GAE GNN baseline actual #739 vs paper #757 18 places worse	0.6639 ±0.0024 best reported	actual #739 vs paper #757 18 places worse paper 0.6580 ±0.0040 best at pub: Decoupled Self-supervised Learning for Non-Homophilou Graphs (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.6580 ±0.0040 Δ at pub 0.0059

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 #585 567 places worse	0.9296 ±0.0015 best reported	actual #18 vs paper #585 567 places worse paper 0.7900 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 0.7900 Δ at pub 0.1104 Δ today 0.1396
20	GAT GNN baseline actual #20 vs paper #585 565 places worse	0.9233 ±0.0018 best reported	actual #20 vs paper #585 565 places worse paper 0.7900 ±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 0.7900 ±0.0030 Δ at pub 0.1070 Δ today 0.1333
116	SAGE GNN global	0.8881 ±0.0015	Efficient Text-Attributed Graph Learning through Selective Annotation and Graph Alignment
136	CCA-SSG GNN baseline actual #136 vs paper #340 204 places worse	0.8840 ±0.0060 best reported	actual #136 vs paper #340 204 places worse paper 0.8160 ±0.0040 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.8160 ±0.0040 Δ at pub 0.0680
144	BGRL GNN baseline actual #144 vs paper #487 343 places worse	0.8830 ±0.0010 best reported	actual #144 vs paper #487 343 places worse paper 0.7960 ±0.0050 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.7960 ±0.0050 Δ at pub 0.0870
154	SUGRL GNN baseline actual #154 vs paper #332 178 places worse	0.8820 ±0.0020 best reported	actual #154 vs paper #332 178 places worse paper 0.8190 ±0.0030 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.8190 ±0.0030 Δ at pub 0.0630
194	GRACE GNN baseline actual #194 vs paper #398 204 places worse	0.8730 ±0.0010 best reported	actual #194 vs paper #398 204 places worse paper 0.8060 ±0.0040 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) paper 0.8060 ±0.0040 Δ at pub 0.0670
196	MVGRL GNN baseline actual #196 vs paper #439 243 places worse	0.8725 ±0.0012 best reported	actual #196 vs paper #439 243 places worse paper 0.8010 ±0.0070 best at pub: Node Representation Learning in Graph via Node-to-Neighbourhood Mutual Information Maximization (2022) paper 0.8010 ±0.0070 Δ at pub 0.0715
204	DGI GNN baseline actual #204 vs paper #745 541 places worse	0.8708 ±0.0017 best reported	actual #204 vs paper #745 541 places worse paper 0.7680 ±0.0060 best at pub: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) paper 0.7680 ±0.0060 Δ at pub 0.1028
208	COSTA GNN baseline actual #208 vs paper #364 156 places worse	0.8698 ±0.0033 best reported	actual #208 vs paper #364 156 places worse paper 0.8110 ±0.0020 best at pub: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) paper 0.8110 ±0.0020 Δ at pub 0.0588
210	MLP code ↗ Trad global	0.8690 ±0.0074	On the Equivalence of Graph Convolution and Mixup code ↗
232	GraphMAE GNN baseline actual #232 vs paper #364 132 places worse	0.8594 ±0.0016 best reported	actual #232 vs paper #364 132 places worse paper 0.8110 ±0.0040 best at pub: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) paper 0.8110 ±0.0040 Δ at pub 0.0484
255	GCA GNN baseline actual #255 vs paper #428 173 places worse	0.8512 ±0.0023 best reported	actual #255 vs paper #428 173 places worse paper 0.8020 ±0.0040 best at pub: LAC: Graph Contrastive Learning with Learnable Augmentation in Continuous Space (2024) paper 0.8020 ±0.0040 Δ at pub 0.0492
299	VGAE GNN baseline actual #299 vs paper #775 476 places worse	0.8307 ±0.0029 best reported	actual #299 vs paper #775 476 places worse paper 0.7580 ±0.0040 best at pub: LAC: Graph Contrastive Learning with Learnable Augmentation in Continuous Space (2024) paper 0.7580 ±0.0040 Δ at pub 0.0727
304	GAE GNN baseline actual #304 vs paper #823 519 places worse	0.8294 ±0.0031 best reported	actual #304 vs paper #823 519 places worse paper 0.7210 ±0.0050 best at pub: LAC: Graph Contrastive Learning with Learnable Augmentation in Continuous Space (2024) paper 0.7210 ±0.0050 Δ at pub 0.1084
313	RelGCL_List ★ proposed	0.8270 ±0.0040	This paper
324	RelGCL_Pair ★ proposed	0.8220 ±0.0030	This paper
330	GraphACL GNN baseline actual #330 vs paper #330	0.8202 ±0.0015 best reported	actual #330 vs paper #330 paper 0.8200 ±0.0020 best today: Discrepancy-Aware Graph Mask Auto-Encoder (KDD 2025) Difference within reported uncertainty; not counted as underperformance. paper 0.8200 ±0.0020 Δ today 0.0002
365	HomoGCL GNN baseline	0.8110 ±0.0030	This paper
380	AutoSSL GNN baseline	0.8090 ±0.0060	This paper
384	AFGRL GNN baseline actual #384 vs paper #384	0.8080 ±0.0060	actual #384 vs paper #384 paper 0.8080 ±0.0060 prev best: HomoGCL: Rethinking Homophily in Graph Contrastive Learning (KDD 2023) Improvement within reported uncertainty; not highlighted. paper 0.8080 ±0.0060 Δ at pub 0.0080
407	gCooL GNN baseline actual #407 vs paper #407	0.8050 ±0.0040	actual #407 vs paper #407 paper 0.8050 ±0.0040 prev best: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) Improvement within reported uncertainty; not highlighted. paper 0.8050 ±0.0040 Δ at pub 0.0030
414	G-BT GNN baseline	0.8040 ±0.0060	This paper
559	SP-GCL GNN baseline actual #559 vs paper #574 15 places worse	0.7920 ±0.0070 best reported	actual #559 vs paper #574 15 places worse paper 0.7910 ±0.0080 best at pub: Graph Contrastive Learning under Heterophily via Graph Filters (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.7910 ±0.0080 Δ at pub 0.0010

Amazon-Photo

Global leaderboard →

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 #269 267 places worse	0.9412 ±0.0008 best reported	actual #2 vs paper #269 267 places worse paper 0.8651 ±0.0054 best at pub: Diss-l-ECT: Dissecting Graph Data with local Euler Characteristic Transforms (ICML 2024) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.8651 ±0.0054 Δ at pub 0.0509 Δ today 0.0761
3	GAT GNN baseline actual #3 vs paper #256 253 places worse	0.9398 ±0.0022 best reported	actual #3 vs paper #256 253 places worse paper 0.8693 ±0.0029 best at pub: Diss-l-ECT: Dissecting Graph Data with local Euler Characteristic Transforms (ICML 2024) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.8693 ±0.0029 Δ at pub 0.0547 Δ today 0.0705
8	SAGE code ↗ GNN global	0.9351 ±0.0006	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
79	BGRL GNN baseline actual #79 vs paper #129 50 places worse	0.9034 ±0.0019 best reported	actual #79 vs paper #129 50 places worse paper 0.8962 ±0.0037 best at pub: Spectral Augmentation for Self-Supervised Learning on Graphs (ICLR 2022) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.8962 ±0.0037 Δ at pub 0.0072 Δ today 0.0072
90	RelGCL_Pair ★ proposed	0.9014 ±0.0035	This paper
100	GraphMAE GNN baseline actual #100 vs paper #432 332 places worse	0.9002 ±0.0024 best reported	actual #100 vs paper #432 332 places worse paper 0.7944 ±0.0048 best at pub: Rethinking and Simplifying Bootstrapped Graph Latents (2023) best today: Discrepancy-Aware Graph Mask Auto-Encoder (KDD 2025) paper 0.7944 ±0.0048 Δ at pub 0.1044 Δ today 0.1058
104	RelGCL_List ★ proposed	0.8999 ±0.0038	This paper
112	AFGRL GNN baseline	0.8988 ±0.0033	This paper
118	GraphACL GNN baseline	0.8980 ±0.0025	This paper
128	SP-GCL GNN baseline	0.8968 ±0.0019	This paper
139	GRACE GNN baseline actual #139 vs paper #282 143 places worse	0.8953 ±0.0035 best reported	actual #139 vs paper #282 143 places worse paper 0.8625 ±0.0025 best at pub: Spectral Augmentation for Self-Supervised Learning on Graphs (ICLR 2022) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.8625 ±0.0025 Δ at pub 0.0328 Δ today 0.0328
157	HomoGCL GNN baseline actual #157 vs paper #186 29 places worse	0.8900 ±0.0010 best reported	actual #157 vs paper #186 29 places worse paper 0.8846 ±0.0020 best at pub: Multiplex Graph Contrastive Learning with Soft Negatives (2024) paper 0.8846 ±0.0020 Δ at pub 0.0054
159	CCA-SSG GNN baseline actual #159 vs paper #171 12 places worse	0.8896 ±0.0013 best reported	actual #159 vs paper #171 12 places worse paper 0.8874 ±0.0028 best at pub: Str-GCL: Structural Commonsense Driven Graph Contrastive Learning (WWW 2025) Difference within reported uncertainty; not counted as underperformance. paper 0.8874 ±0.0028 Δ at pub 0.0022
160	GCA GNN baseline actual #160 vs paper #213 53 places worse	0.8894 ±0.0015 best reported	actual #160 vs paper #213 53 places worse paper 0.8785 ±0.0031 best at pub: Barlow Graph Auto-Encoder for Unsupervised Network Embedding (2021) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.8785 ±0.0031 Δ at pub 0.0109 Δ today 0.0109
162	SUGRL GNN baseline actual #162 vs paper #162	0.8893 ±0.0021 best reported	actual #162 vs paper #162 paper 0.8891 ±0.0022 best at pub: Localized Graph Contrastive Learning (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.8891 ±0.0022 Δ at pub 0.0002
166	gCooL GNN baseline	0.8885 ±0.0014	This paper
179	MVGRL GNN baseline actual #179 vs paper #235 56 places worse	0.8861 ±0.0064 best reported	actual #179 vs paper #235 56 places worse paper 0.8752 ±0.0011 best at pub: Balancing Graph Embedding Smoothness in Self-Supervised Learning via Information-Theoretic Decomposition (WWW 2025) paper 0.8752 ±0.0011 Δ at pub 0.0109
188	AutoSSL GNN baseline actual #188 vs paper #200 12 places worse	0.8840 ±0.0030 best reported	actual #188 vs paper #200 12 places worse paper 0.8818 ±0.0043 best at pub: Decoupling Weighing and Selecting for Integrating Multiple Graph Pre-training Tasks (ICLR 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8818 ±0.0043 Δ at pub 0.0022
192	COSTA GNN baseline actual #192 vs paper #193 1 place worse	0.8837 ±0.0022 best reported	actual #192 vs paper #193 1 place worse paper 0.8832 ±0.0003 best at pub: HomoGCL: Rethinking Homophily in Graph Contrastive Learning (KDD 2023) Difference within reported uncertainty; not counted as underperformance. paper 0.8832 ±0.0003 Δ at pub 0.0005
202	G-BT GNN baseline	0.8814 ±0.0033	This paper
210	DGI GNN baseline actual #210 vs paper #354 144 places worse	0.8798 ±0.0081 best reported	actual #210 vs paper #354 144 places worse paper 0.8395 ±0.0047 best at pub: Bootstrap Latents of Nodes and Neighbors for Graph Self-Supervised Learning (2024) paper 0.8395 ±0.0047 Δ at pub 0.0403
246	VGAE GNN baseline actual #246 vs paper #274 28 places worse	0.8721 ±0.0026 best reported	actual #246 vs paper #274 28 places worse paper 0.8637 ±0.0021 best at pub: HC-GAE: The Hierarchical Cluster-based Graph Auto-Encoder for Graph Representation Learning (NeurIPS 2024) paper 0.8637 ±0.0021 Δ at pub 0.0084
264	GAE GNN baseline actual #264 vs paper #305 41 places worse	0.8666 ±0.0007 best reported	actual #264 vs paper #305 41 places worse paper 0.8527 ±0.0019 best at pub: Graph Masked Autoencoders with Transformers (2022) paper 0.8527 ±0.0019 Δ at pub 0.0139
320	MLP code ↗ Trad global	0.8463	GraphFM: A generalist graph transformer that learns transferable representations across diverse … code ↗

WikiCS

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	GCN GNN baseline actual #1 vs paper #202 201 places worse	0.9302 ±0.0011 best reported	actual #1 vs paper #202 201 places worse paper 0.7719 ±0.0012 best at pub: Augmentation-Free Graph Contrastive Learning with Performance Guarantee (2022) paper 0.7719 ±0.0012 Δ at pub 0.1583
2	RSE + RGCN (70b) ext data global	0.8852 ±0.0019	Unleashing the Potential of Text-attributed Graphs: Automatic Relation Decomposition via Large …
3	LEMP+T ext data Hyb global	0.8825	Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
9	SAGE GNN global	0.8771	Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
31	GAT GNN baseline actual #31 vs paper #187 156 places worse	0.8533 best reported	actual #31 vs paper #187 156 places worse paper 0.7765 ±0.0011 best at pub: GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers (ICML 2024) best today: Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning (2025) paper 0.7765 ±0.0011 Δ at pub 0.0557 Δ today 0.0768
84	GraphMAE GNN baseline actual #84 vs paper #205 121 places worse	0.8052 ±0.0032 best reported	actual #84 vs paper #205 121 places worse paper 0.7712 ±0.0030 best at pub: Rethinking and Simplifying Bootstrapped Graph Latents (2023) best today: Discrepancy-Aware Graph Mask Auto-Encoder (KDD 2025) paper 0.7712 ±0.0030 Δ at pub 0.0280 Δ today 0.0340
93	RelGCL_List ★ proposed	0.8016 ±0.0058	This paper
96	RelGCL_Pair ★ proposed	0.8014 ±0.0051	This paper
94	GRACE GNN baseline actual #94 vs paper #171 77 places worse	0.8014 ±0.0048 best reported	actual #94 vs paper #171 77 places worse paper 0.7819 ±0.0001 best at pub: Large-Scale Representation Learning on Graphs via Bootstrapping (ICLR 2021) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.7819 ±0.0001 Δ at pub 0.0195 Δ today 0.0195
100	BGRL GNN baseline actual #100 vs paper #117 17 places worse	0.8000 ±0.0010 best reported	actual #100 vs paper #117 17 places worse paper 0.7931 ±0.0055 best at pub: Learning Graph Augmentations to Learn Graph Representations (2022) best today: PyG-SSL: A Graph Self-Supervised Learning Toolkit (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7931 ±0.0055 Δ at pub 0.0069 Δ today 0.0069
108	GraphACL GNN baseline actual #108 vs paper #118 10 places worse	0.7966 ±0.0029 best reported	actual #108 vs paper #118 10 places worse paper 0.7927 ±0.0045 best today: Discrepancy-Aware Graph Mask Auto-Encoder (KDD 2025) Difference within reported uncertainty; not counted as underperformance. paper 0.7927 ±0.0045 Δ today 0.0039
129	COSTA GNN baseline	0.7912 ±0.0002	This paper
130	CCA-SSG GNN baseline actual #130 vs paper #150 20 places worse	0.7908 ±0.0053 best reported	actual #130 vs paper #150 20 places worse paper 0.7865 ±0.0014 best at pub: Rethinking and Simplifying Bootstrapped Graph Latents (2023) best today: Bootstrap Latents of Nodes and Neighbors for Graph Self-Supervised Learning (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7865 ±0.0014 Δ at pub 0.0043 Δ today 0.0043
135	SP-GCL GNN baseline	0.7901 ±0.0051	This paper
138	SUGRL GNN baseline	0.7897 ±0.0022	This paper
143	AFGRL GNN baseline actual #143 vs paper #187 44 places worse	0.7890 ±0.0060 best reported	actual #143 vs paper #187 44 places worse paper 0.7762 ±0.0049 best at pub: PyG-SSL: A Graph Self-Supervised Learning Toolkit (2024) paper 0.7762 ±0.0049 Δ at pub 0.0128
145	HomoGCL GNN baseline actual #145 vs paper #145	0.7884 ±0.0047	actual #145 vs paper #145 paper 0.7884 ±0.0047 prev best: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) Improvement within reported uncertainty; not highlighted. paper 0.7884 ±0.0047 Δ at pub 0.0058
147	gCooL GNN baseline	0.7874 ±0.0004	This paper
151	DGI GNN baseline actual #151 vs paper #228 77 places worse	0.7865 ±0.0690 best reported	actual #151 vs paper #228 77 places worse paper 0.7535 ±0.0014 best at pub: Bootstrap Latents of Nodes and Neighbors for Graph Self-Supervised Learning (2024) best today: Quantizing Text-attributed Graphs for Semantic-Structural Integration (KDD 2025) paper 0.7535 ±0.0014 Δ at pub 0.0290 Δ today 0.0330
162	GCA GNN baseline actual #162 vs paper #166 4 places worse	0.7840 ±0.0013 best reported	actual #162 vs paper #166 4 places worse paper 0.7830 best at pub: Learning Graph Augmentations to Learn Graph Representations (2022) best today: Augmentation-Free Graph Contrastive Learning with Performance Guarantee (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.7830 Δ at pub 0.0010 Δ today 0.0010
181	GAE GNN baseline actual #181 vs paper #265 84 places worse	0.7787 ±0.0053 best reported	actual #181 vs paper #265 84 places worse paper 0.7015 ±0.0001 best at pub: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) paper 0.7015 ±0.0001 Δ at pub 0.0772
180	VGAE GNN baseline actual #180 vs paper #222 42 places worse	0.7787 ±0.0053 best reported	actual #180 vs paper #222 42 places worse paper 0.7563 ±0.0019 best at pub: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) paper 0.7563 ±0.0019 Δ at pub 0.0224
190	MVGRL GNN baseline actual #190 vs paper #191 1 place worse	0.7757 ±0.0046 best reported	actual #190 vs paper #191 1 place worse paper 0.7752 ±0.0008 best at pub: Bootstrap Latents of Nodes and Neighbors for Graph Self-Supervised Learning (2024) best today: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7752 ±0.0008 Δ at pub 0.0005 Δ today 0.0005
211	G-BT GNN baseline actual #211 vs paper #212 1 place worse	0.7683 ±0.0073 best reported	actual #211 vs paper #212 1 place worse paper 0.7665 ±0.0062 best at pub: Bootstrap Latents of Nodes and Neighbors for Graph Self-Supervised Learning (2024) best today: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.7665 ±0.0062 Δ at pub 0.0018 Δ today 0.0018
212	AutoSSL GNN baseline	0.7680 ±0.0013	This paper
237	MLP Trad global	0.7469 ±0.0085	View Space: Learning Representation across Arbitrary Graphs

Coauthor-CS

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	DoG code ↗ GNN global	0.9690 ±0.0030	Diffusion on Graph: Augmentation of Graph Structure for Node Classification code ↗
2	GWN-fa code ↗ Hyb global	0.9667 ±0.0026	Graph Wave Networks code ↗
3	GWN-sym code ↗ GNN global	0.9666 ±0.0026	Graph Wave Networks code ↗
6	GraphSAGE code ↗ GNN global	0.9593 ±0.0025	Graph Wave Networks code ↗
12	GCN GNN baseline actual #12 vs paper #67 55 places worse	0.9514 ±0.0025 best reported	actual #12 vs paper #67 55 places worse paper 0.9303 ±0.0031 best at pub: Graph Wave Networks (WWW 2025) paper 0.9303 ±0.0031 Δ at pub 0.0211
21	GAT GNN baseline actual #21 vs paper #107 86 places worse	0.9437 ±0.0028 best reported	actual #21 vs paper #107 86 places worse paper 0.9231 ±0.0024 best at pub: Graph Wave Networks (WWW 2025) paper 0.9231 ±0.0024 Δ at pub 0.0206
41	RelGCL_Pair ★ proposed	0.9353 ±0.0012	This paper
44	RelGCL_List ★ proposed	0.9350 ±0.0012	This paper
43	DGI GNN baseline actual #43 vs paper #114 71 places worse	0.9350 ±0.0030 best reported	actual #43 vs paper #114 71 places worse paper 0.9215 ±0.0063 best at pub: PyG-SSL: A Graph Self-Supervised Learning Toolkit (2024) paper 0.9215 ±0.0063 Δ at pub 0.0135
48	CCA-SSG GNN baseline actual #48 vs paper #52 4 places worse	0.9337 ±0.0020 best reported	actual #48 vs paper #52 4 places worse paper 0.9331 ±0.0022 best at pub: MA-GCL: Model Augmentation Tricks for Graph Contrastive Learning (AAAI 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.9331 ±0.0022 Δ at pub 0.0006
50	AutoSSL GNN baseline	0.9335 ±0.0009	This paper
51	GCA GNN baseline actual #51 vs paper #63 12 places worse	0.9332 ±0.0012 best reported	actual #51 vs paper #63 12 places worse paper 0.9310 ±0.0001 best at pub: Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating (AAAI 2022) paper 0.9310 ±0.0001 Δ at pub 0.0022
52	gCooL GNN baseline actual #52 vs paper #52	0.9332 ±0.0002	actual #52 vs paper #52 paper 0.9332 ±0.0002 prev best: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) paper 0.9332 ±0.0002 Δ at pub 0.0057
53	BGRL GNN baseline actual #53 vs paper #98 45 places worse	0.9331 ±0.0013 best reported	actual #53 vs paper #98 45 places worse paper 0.9267 ±0.0021 best at pub: Spectral Augmentation for Self-Supervised Learning on Graphs (ICLR 2022) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.9267 ±0.0021 Δ at pub 0.0064 Δ today 0.0064
56	AFGRL GNN baseline actual #56 vs paper #58 2 places worse	0.9330 ±0.0020 best reported	actual #56 vs paper #58 2 places worse paper 0.9327 ±0.0017 best at pub: PyG-SSL: A Graph Self-Supervised Learning Toolkit (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9327 ±0.0017 Δ at pub 0.0003
63	GraphMAE GNN baseline actual #63 vs paper #63	0.9313 ±0.0015	actual #63 vs paper #63 paper 0.9313 ±0.0015 prev best: Diffusion on Graph: Augmentation of Graph Structure for Node Classification (TMLR 2025) Improvement within reported uncertainty; not highlighted. paper 0.9313 ±0.0015 Δ at pub 0.0013
69	GRACE GNN baseline actual #69 vs paper #82 13 places worse	0.9301 ±0.0003 best reported	actual #69 vs paper #82 13 places worse paper 0.9293 ±0.0001 best at pub: A Fused Gromov-Wasserstein Approach to Subgraph Contrastive Learning (TMLR 2025) paper 0.9293 ±0.0001 Δ at pub 0.0008
77	G-BT GNN baseline	0.9295 ±0.0017	This paper
76	COSTA GNN baseline actual #76 vs paper #80 4 places worse	0.9295 ±0.0012 best reported	actual #76 vs paper #80 4 places worse paper 0.9294 ±0.0010 best at pub: Rethinking and Simplifying Bootstrapped Graph Latents (2023) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.9294 ±0.0010 Δ at pub 0.0001 Δ today 0.0001
83	SUGRL GNN baseline actual #83 vs paper #87 4 places worse	0.9290 ±0.0010 best reported	actual #83 vs paper #87 4 places worse paper 0.9283 ±0.0023 best at pub: PyG-SSL: A Graph Self-Supervised Learning Toolkit (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9283 ±0.0023 Δ at pub 0.0007
88	MLP Trad global	0.9280 ±0.0020	Revisiting the Necessity of Graph Learning and Common Graph Benchmarks
93	GraphACL GNN baseline	0.9277 ±0.0014	This paper
94	HomoGCL GNN baseline actual #94 vs paper #94	0.9274 ±0.0022	actual #94 vs paper #94 paper 0.9274 ±0.0022 prev best: Negative-Free Self-Supervised Gaussian Embedding of Graphs (2024) paper 0.9274 ±0.0022 Δ at pub 0.0046
117	VGAE GNN baseline	0.9211 ±0.0009	This paper
116	MVGRL GNN baseline	0.9211 ±0.0012	This paper
127	SP-GCL GNN baseline	0.9192 ±0.0010	This paper
163	GAE GNN baseline actual #163 vs paper #163	0.9005 ±0.0035 best reported	actual #163 vs paper #163 paper 0.9001 ±0.0071 best at pub: HAT-GAE: Self-Supervised Graph Auto-encoders with Hierarchical Adaptive Masking and Trainable Corruption (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.9001 ±0.0071 Δ at pub 0.0004

Heterophilic Graphs

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 ↗
42	MLP Trad global	0.4026 ±0.0108	Heterophily-Aware Graph Attention Network
68	GraphSAGE code ↗ GNN global	0.3899 ±0.0085	A Quasi-Wasserstein Loss for Learning Graph Neural Networks code ↗
251	GRACE GNN baseline actual #251 vs paper #390 139 places worse	0.3450 ±0.0110 best reported	actual #251 vs paper #390 139 places worse paper 0.2827 ±0.0043 best at pub: Graph Contrastive Learning under Heterophily via Graph Filters (2023) paper 0.2827 ±0.0043 Δ at pub 0.0623
335	DGI GNN baseline actual #335 vs paper #389 54 places worse	0.3067 ±0.0063 best reported	actual #335 vs paper #389 54 places worse paper 0.2830 ±0.0076 best at pub: Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating (AAAI 2022) best today: Toward General Digraph Contrastive Learning: A Dual Spatial Perspective (2025) paper 0.2830 ±0.0076 Δ at pub 0.0152 Δ today 0.0237
342	RelGCL_Pair ★ proposed	0.3032 ±0.0091	This paper
345	RelGCL_List ★ proposed	0.3023 ±0.0088	This paper
351	GraphACL GNN baseline	0.3003 ±0.0013	This paper
356	BGRL GNN baseline actual #356 vs paper #377 21 places worse	0.2986 best reported	actual #356 vs paper #377 21 places worse paper 0.2880 ±0.0054 best at pub: Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating (AAAI 2022) paper 0.2880 ±0.0054 Δ at pub 0.0106
361	GCA GNN baseline actual #361 vs paper #376 15 places worse	0.2965 best reported	actual #361 vs paper #376 15 places worse paper 0.2889 ±0.0050 best at pub: Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating (AAAI 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.2889 ±0.0050 Δ at pub 0.0076
376	SP-GCL GNN baseline actual #376 vs paper #376	0.2894 ±0.0069	actual #376 vs paper #376 paper 0.2894 ±0.0069 prev best: Graph Contrastive Learning under Heterophily via Graph Filters (2023) paper 0.2894 ±0.0069 Δ at pub 0.0124
407	AFGRL GNN baseline	0.2743 ±0.0131	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 ↗
52	GCN GNN global	0.7033	Understanding Heterophily for Graph Neural Networks
58	RelGCL_Pair ★ proposed	0.6925 ±0.0089	This paper
62	RelGCL_List ★ proposed	0.6906 ±0.0086	This paper
90	SP-GCL GNN baseline	0.6528 ±0.0053	This paper
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
46	GCN GNN global	0.6067	Understanding Heterophily for Graph Neural Networks
57	RelGCL_Pair ★ proposed	0.5767 ±0.0096	This paper
58	RelGCL_List ★ proposed	0.5753 ±0.0092	This paper
77	GraphACL GNN baseline	0.5405 ±0.0013	This paper
84	SP-GCL GNN baseline	0.5210 ±0.0067	This paper
98	GCA GNN baseline actual #98 vs paper #150 52 places worse	0.4809 ±0.0021 best reported	actual #98 vs paper #150 52 places worse paper 0.4153 ±0.0109 best at pub: Spectral-Aware Augmentation for Enhanced Graph Representation Learning (2023) best today: Discrepancy-Aware Graph Mask Auto-Encoder (KDD 2025) paper 0.4153 ±0.0109 Δ at pub 0.0069 Δ today 0.0656
108	BGRL GNN baseline	0.4624 ±0.0070	This paper
142	AFGRL GNN baseline	0.4236 ±0.0040	This paper
144	DGI GNN baseline	0.4222 ±0.0063	This paper
154	GRACE GNN baseline	0.4109 ±0.0085	This paper
172	MLP code ↗ Trad global	0.3930 ±0.0079	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗

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 Improved at pub — later paper surpassed this result Best today — OK at publication; today's leaderboard best is elsewhere

#	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 #275 273 places worse	0.9669 ±0.0014 best reported	actual #2 vs paper #275 273 places worse paper 0.9256 ±0.0035 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 0.9256 ±0.0035 Δ at pub 0.0242 Δ today 0.0413
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 #292 284 places worse	0.9632 ±0.0008 best reported	actual #8 vs paper #292 284 places worse paper 0.9242 ±0.0022 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 0.9242 ±0.0022 Δ at pub 0.0234 Δ today 0.0390
130	AFGRL GNN baseline actual #130 vs paper #189 59 places worse	0.9361 ±0.0020 best reported	actual #130 vs paper #189 59 places worse paper 0.9322 ±0.0028 best at pub: Wiener Graph Deconvolutional Network Improves Graph Self-Supervised Learning (AAAI 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.9322 ±0.0028 Δ at pub 0.0039
153	AutoSSL GNN baseline actual #153 vs paper #254 101 places worse	0.9350 ±0.0070 best reported	actual #153 vs paper #254 101 places worse paper 0.9271 ±0.0032 best at pub: Decoupling Weighing and Selecting for Integrating Multiple Graph Pre-training Tasks (ICLR 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9271 ±0.0032 Δ at pub 0.0079
163	RelGCL_Pair ★ proposed	0.9342 ±0.0044	This paper
164	GraphMAE GNN baseline actual #164 vs paper #415 251 places worse	0.9341 ±0.0010 best reported	actual #164 vs paper #415 251 places worse paper 0.9071 ±0.0040 best at pub: Rethinking and Simplifying Bootstrapped Graph Latents (2023) paper 0.9071 ±0.0040 Δ at pub 0.0270
171	RelGCL_List ★ proposed	0.9340 ±0.0048	This paper
172	BGRL GNN baseline actual #172 vs paper #214 42 places worse	0.9337 ±0.0021 best reported	actual #172 vs paper #214 42 places worse paper 0.9307 ±0.0034 best at pub: Wiener Graph Deconvolutional Network Improves Graph Self-Supervised Learning (AAAI 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.9307 ±0.0034 Δ at pub 0.0030
175	GraphACL GNN baseline	0.9331 ±0.0019	This paper
186	CCA-SSG GNN baseline actual #186 vs paper #201 15 places worse	0.9325 ±0.0021 best reported	actual #186 vs paper #201 15 places worse paper 0.9314 ±0.0014 best at pub: Wiener Graph Deconvolutional Network Improves Graph Self-Supervised Learning (AAAI 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.9314 ±0.0014 Δ at pub 0.0011
191	GCA GNN baseline actual #191 vs paper #283 92 places worse	0.9320 ±0.0060 best reported	actual #191 vs paper #283 92 places worse paper 0.9249 ±0.0009 best at pub: Unifying Graph Contrastive Learning with Flexible Contextual Scopes (2022) best today: Decoupling Weighing and Selecting for Integrating Multiple Graph Pre-training Tasks (ICLR 2024) paper 0.9249 ±0.0009 Δ at pub 0.0071 Δ today 0.0071
194	SUGRL GNN baseline actual #194 vs paper #233 39 places worse	0.9320 ±0.0020 best reported	actual #194 vs paper #233 39 places worse paper 0.9285 ±0.0024 best at pub: Oversmoothing: A Nightmare for Graph Contrastive Learning? (2023) best today: PyG-SSL: A Graph Self-Supervised Learning Toolkit (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9285 ±0.0024 Δ at pub 0.0035 Δ today 0.0035
198	gCooL GNN baseline	0.9318 ±0.0012	This paper
203	GRACE GNN baseline actual #203 vs paper #324 121 places worse	0.9313 ±0.0017 best reported	actual #203 vs paper #324 121 places worse paper 0.9215 ±0.0024 best at pub: Learning Graph Augmentations to Learn Graph Representations (2022) best today: Learn Beneficial Noise as Graph Augmentation (ICML 2025) paper 0.9215 ±0.0024 Δ at pub 0.0065 Δ today 0.0098
206	MVGRL GNN baseline actual #206 vs paper #359 153 places worse	0.9310 ±0.0040 best reported	actual #206 vs paper #359 153 places worse paper 0.9174 ±0.0007 best at pub: Decoupling Weighing and Selecting for Integrating Multiple Graph Pre-training Tasks (ICLR 2024) paper 0.9174 ±0.0007 Δ at pub 0.0136
213	DGI GNN baseline actual #213 vs paper #366 153 places worse	0.9309 ±0.0008 best reported	actual #213 vs paper #366 153 places worse paper 0.9161 ±0.0022 best at pub: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) paper 0.9161 ±0.0022 Δ at pub 0.0148
222	HomoGCL GNN baseline actual #222 vs paper #225 3 places worse	0.9300 ±0.0020 best reported	actual #222 vs paper #225 3 places worse paper 0.9292 ±0.0018 best at pub: Multiplex Graph Contrastive Learning with Soft Negatives (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9292 ±0.0018 Δ at pub 0.0008
263	COSTA GNN baseline actual #263 vs paper #275 12 places worse	0.9267 ±0.0016 best reported	actual #263 vs paper #275 12 places worse paper 0.9256 ±0.0045 best at pub: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9256 ±0.0045 Δ at pub 0.0011
267	G-BT GNN baseline	0.9263 ±0.0044	This paper
287	SP-GCL GNN baseline	0.9249 ±0.0031	This paper
317	VGAE GNN baseline actual #317 vs paper #324 7 places worse	0.9224 best reported	actual #317 vs paper #324 7 places worse paper 0.9220 ±0.0011 best at pub: Barlow Graph Auto-Encoder for Unsupervised Network Embedding (2021) Difference within reported uncertainty; not counted as underperformance. paper 0.9220 ±0.0011 Δ at pub 0.0004
352	GAE GNN baseline actual #352 vs paper #366 14 places worse	0.9185 ±0.0006 best reported	actual #352 vs paper #366 14 places worse paper 0.9162 ±0.0013 best at pub: Graph Masked Autoencoders with Transformers (2022) paper 0.9162 ±0.0013 Δ at pub 0.0023
390	MLP code ↗ Trad global	0.9128 ±0.0063	Clarify Confused Nodes via Separated Learning code ↗