Calibrating and Improving Graph Contrastive Learning

Kaili Ma, Haochen Yang, Han Yang, Tatiana Jin, Pengfei Chen, Yongqiang Chen, Barakeel Fanseu Kamhoua, James Cheng

Abstract

Graph contrastive learning algorithms have demonstrated remarkable success in various applications such as node classification, link prediction, and graph clustering.

Benchmark Results

Tested on 5 datasets 10 baseline values reported too low

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

Classic

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 #552 534 places worse	0.9296 ±0.0015 best reported	actual #18 vs paper #552 534 places worse paper 0.7926 ±0.0038 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 0.7926 ±0.0038 Δ at pub 0.1044 Δ today 0.1370
116	SAGE GNN global	0.8881 ±0.0015	Efficient Text-Attributed Graph Learning through Selective Annotation and Graph Alignment
195	GMI GNN baseline actual #195 vs paper #682 487 places worse	0.8726 ±0.0016 best reported	actual #195 vs paper #682 487 places worse paper 0.7797 ±0.0104 best at pub: SAIL: Self-Augmented Graph Contrastive Learning (AAAI 2020) best today: Node Representation Learning in Graph via Node-to-Neighbourhood Mutual Information Maximization (2022) paper 0.7797 ±0.0104 Δ at pub 0.0213 Δ today 0.0929
204	DGI GNN baseline actual #204 vs paper #743 539 places worse	0.8708 ±0.0017 best reported	actual #204 vs paper #743 539 places worse paper 0.7689 ±0.0053 best at pub: Contrastive and Generative Graph Convolutional Networks for Graph-based Semi-Supervised Learning (AAAI 2020) best today: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) paper 0.7689 ±0.0053 Δ at pub 0.0041 Δ today 0.1019
210	MLP code ↗ Trad global	0.8690 ±0.0074	On the Equivalence of Graph Convolution and Mixup code ↗
251	Node2Vec Walk baseline actual #251 vs paper #858 607 places worse	0.8520 ±0.0002 best reported	actual #251 vs paper #858 607 places worse paper 0.6634 ±0.0140 best at pub: Attention-based Graph Neural Network for Semi-supervised Learning (2018) best today: Node Representation Learning in Graph via Node-to-Neighbourhood Mutual Information Maximization (2022) paper 0.6634 ±0.0140 Δ at pub 0.0896 Δ today 0.1886

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 code ↗ GNN global	0.9669 ±0.0014	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
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 #479 471 places worse	0.9632 ±0.0008 best reported	actual #8 vs paper #479 471 places worse paper 0.8875 ±0.0199 best at pub: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.8875 ±0.0199 Δ at pub 0.0367 Δ today 0.0757
164	GraphMAE ext data 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 ±0.0129 best today: Rethinking and Simplifying Bootstrapped Graph Latents (2023) paper 0.9071 ±0.0129 Δ today 0.0270
191	GCA ext data GNN baseline actual #191 vs paper #202 11 places worse	0.9320 ±0.0060 best reported	actual #191 vs paper #202 11 places worse paper 0.9313 ±0.0032 best today: Decoupling Weighing and Selecting for Integrating Multiple Graph Pre-training Tasks (ICLR 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9313 ±0.0032 Δ today 0.0007
213	DGI GNN baseline actual #213 vs paper #544 331 places worse	0.9309 ±0.0008 best reported	actual #213 vs paper #544 331 places worse paper 0.7719 ±0.0260 best at pub: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) paper 0.7719 ±0.0260 Δ at pub 0.1442 Δ today 0.1590
390	MLP code ↗ Trad global	0.9128 ±0.0063	Clarify Confused Nodes via Separated Learning code ↗
408	GMI GNN baseline actual #408 vs paper #497 89 places worse	0.9081 ±0.0015 best reported	actual #408 vs paper #497 89 places worse paper 0.8708 ±0.0123 best at pub: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Energy-Balanced Hyperspherical Graph Representation Learning via Structural Binding and Entropic Dispersion (2025) paper 0.8708 ±0.0123 Δ at pub 0.0360 Δ today 0.0373
457	Node2Vec Walk baseline actual #457 vs paper #521 64 places worse	0.8970 ±0.0010 best reported	actual #457 vs paper #521 64 places worse paper 0.8363 ±0.0153 best at pub: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Multiplex Graph Contrastive Learning with Soft Negatives (2024) paper 0.8363 ±0.0153 Δ at pub 0.0604 Δ today 0.0607
494	LC ext data GNN baseline	0.8759 ±0.0150	This paper
499	ML ext data GNN baseline	0.8678 ±0.0170	This paper

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 #430 428 places worse	0.9412 ±0.0008 best reported	actual #2 vs paper #430 428 places worse paper 0.7982 ±0.0204 best at pub: Learning on Attribute-Missing Graphs (2020) best today: Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification (2026) paper 0.7982 ±0.0204 Δ at pub 0.0803 Δ today 0.1430
3	GAT code ↗ GNN global	0.9398 ±0.0022	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
8	SAGE code ↗ GNN global	0.9351 ±0.0006	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
100	GraphMAE GNN baseline actual #100 vs paper #417 317 places worse	0.9002 ±0.0024 best reported	actual #100 vs paper #417 317 places worse paper 0.8116 ±0.0110 best today: Discrepancy-Aware Graph Mask Auto-Encoder (KDD 2025) paper 0.8116 ±0.0110 Δ today 0.0886
160	GCA GNN baseline actual #160 vs paper #199 39 places worse	0.8894 ±0.0015 best reported	actual #160 vs paper #199 39 places worse paper 0.8821 ±0.0037 best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.8821 ±0.0037 Δ today 0.0073
210	DGI GNN baseline actual #210 vs paper #479 269 places worse	0.8798 ±0.0081 best reported	actual #210 vs paper #479 269 places worse paper 0.6492 ±0.0193 best at pub: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Bootstrap Latents of Nodes and Neighbors for Graph Self-Supervised Learning (2024) paper 0.6492 ±0.0193 Δ at pub 0.1903 Δ today 0.2306
320	MLP code ↗ Trad global	0.8463	GraphFM: A generalist graph transformer that learns transferable representations across diverse … code ↗
330	Node2Vec Walk baseline actual #330 vs paper #451 121 places worse	0.8440 ±0.0010 best reported	actual #330 vs paper #451 121 places worse paper 0.7537 ±0.0152 best at pub: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Multiplex Graph Contrastive Learning with Soft Negatives (2024) paper 0.7537 ±0.0152 Δ at pub 0.0902 Δ today 0.0903
342	GMI GNN baseline actual #342 vs paper #432 90 places worse	0.8422 ±0.0052 best reported	actual #342 vs paper #432 90 places worse paper 0.7927 ±0.0164 best at pub: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Localized Graph Contrastive Learning (2022) paper 0.7927 ±0.0164 Δ at pub 0.0294 Δ today 0.0495
400	ML GNN baseline	0.8211 ±0.0147	This paper
407	LC GNN baseline	0.8198 ±0.0152	This paper

Global leaderboard →

Default metric is F1. The paper reports 4 values on F1 and 2 values on Accuracy .

#	Model	F1 ↑ default	Accuracy ↑	Paper
1	EnGCN code ↗ Hyb global	0.9714 ± 0.0003	—	Accuracy — A Comprehensive Study on Large-Scale Graph Training: Benchmarking and Rethinking code ↗
2	SAGN+SLE code ↗ GNN global	0.9710 ± 0.0000	—	Accuracy — A Comprehensive Study on Large-Scale Graph Training: Benchmarking and Rethinking code ↗
3	GAMLP(JK) code ↗ Hyb global	0.9704 ± 0.0001	—	Accuracy — Graph Attention Multi-Layer Perceptron code ↗
14	GraphSAGE code ↗ GNN global	0.9660	—	Accuracy — Accelerating Large Scale Real-Time GNN Inference using Channel Pruning code ↗
58	GCN GNN baseline	0.9648 best reported	0.9402 ± 0.0005	Accuracy 0.9402 ± 0.0005 paper 0.9402 ±0.0005 best at pub: DropEdge: Towards Deep Graph Convolutional Networks on Node Classification (ICLR 2019) paper 0.9402 ±0.0005 Δ at pub 0.0246
70	GMI GNN baseline	0.9500 best reported	—	Accuracy — paper 0.9419 ±0.0004 best at pub: Sub-graph Contrast for Scalable Self-Supervised Graph Representation Learning (2020) best today: Generative Subgraph Contrast for Self-Supervised Graph Representation Learning (2022) paper 0.9419 ±0.0004 Δ at pub 0.0081 Δ today 0.0081
77	LC ext data GNN baseline	0.9443 ± 0.0003	—	Accuracy — This paper
78	ML ext data GNN baseline	0.9438 ± 0.0004	—	Accuracy — This paper
82	DGI GNN baseline	0.9414 ±0.0003	—	Accuracy — paper 0.9414 ±0.0003 prev best: Deep Learning on Graphs: A Survey (2018) Improvement within reported uncertainty; not highlighted. paper 0.9414 ±0.0003 Δ at pub 0.0014
145	Node2Vec Walk baseline	0.9326 ±0.0004	0.9326 ± 0.0004	Accuracy 0.9326 ± 0.0004 paper 0.9326 ±0.0004 prev best: Graph Few-shot Learning via Knowledge Transfer (AAAI 2019) paper 0.9326 ±0.0004 Δ at pub 0.4969

OGB

ogbn-products

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	Jacobi GNN global	0.8968	Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
2	GPR GNN global	0.8953	Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
3	LEMP4HG ext data Hyb global	0.8939	Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
4	SAGE GNN global	0.8933	Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning
6	GCN GNN baseline actual #6 vs paper #250 244 places worse	0.8868 best reported	actual #6 vs paper #250 244 places worse paper 0.7564 ±0.0021 best today: Language Models as Messengers: Enhancing Message Passing in Heterophilic Graph Learning (2025) paper 0.7564 ±0.0021 Δ today 0.1304
23	LC ext data GNN baseline	0.8645 ±0.0013	This paper
26	DGI GNN baseline	0.8637 ±0.0019	This paper
30	GraphMAE ext data GNN baseline actual #30 vs paper #324 294 places worse	0.8630 ±0.0007 best reported	actual #30 vs paper #324 294 places worse paper 0.7156 ±0.0019 best today: Towards Graph Foundation Models: Learning Generalities Across Graphs via Task-Trees (ICML 2024) paper 0.7156 ±0.0019 Δ today 0.1474
31	ML ext data GNN baseline	0.8628 ±0.0020	This paper
33	GMI GNN baseline	0.8614 ±0.0016	This paper
48	No pretraining baseline	0.8469 ±0.0079	This paper
313	Node2Vec Walk baseline	0.7249 ±0.0010	This paper
351	MLP Trad global	0.6605 ±0.0020	Latent Graph Inference using Product Manifolds

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

Abstract

Benchmark Results

Classic

PubMed

Amazon-Photo

Amazon-Computers

Reddit

OGB

ogbn-products