Decoupling Weighing and Selecting for Integrating Multiple Graph Pre-training Tasks

Tianyu Fan, Lirong Wu, Yufei Huang, Haitao Lin, Cheng Tan, Zhangyang Gao, Stan Z. Li

Abstract

Recent years have witnessed the great success of graph pre-training for graph representation learning. With hundreds of graph pre-training tasks proposed, integrating knowledge acquired from multiple pre-training tasks has become a popular research topic. In this paper, we identify two important collaborative processes for this topic: (1) select: how to select an optimal task combination from a given task pool based on their compatibility, and (2) weigh: how to weigh the selected tasks based on their importance. While there currently has been a lot of work focused on weighing, comparatively little effort has been devoted to selecting. This paper proposes a novel instance-level framework for integrating multiple graph pre-training tasks, eigh nd elect (WAS), where the two collaborative processes, weighing and selecting, are combined by decoupled siamese networks. Specifically, it first adaptively learns an optimal combination of tasks for each instance from a given task pool, based on which a customized instance-level task weighing strategy is learned. Extensive experiments on 16 graph datasets across node-level and graph-level downstream tasks have demonstrated that by combining a few simple but classical tasks, WAS can achieve comparable performance to other leading counterparts. The code is available at https://github.com/TianyuFan0504/WAS.

Benchmark Results

Tested on 9 datasets 53 baseline values reported too low

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

Classic

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 ↗
136	WAS Hyb ★ proposed	0.9360 ±0.0070	This paper
153	AutoSSL GNN baseline actual #153 vs paper #153	0.9350 ±0.0070	actual #153 vs paper #153 paper 0.9350 ±0.0070 prev best: Graph Soft-Contrastive Learning via Neighborhood Ranking (2022) Improvement within reported uncertainty; not highlighted. paper 0.9350 ±0.0070 Δ at pub 0.0079
169	ParetoGNN GNN baseline	0.9340 ±0.0020	This paper
181	AGSSL GNN baseline	0.9330 ±0.0040	This paper
191	GCA GNN baseline	0.9320 ±0.0060	This paper
203	GRACE GNN baseline actual #203 vs paper #294 91 places worse	0.9313 ±0.0017 best reported	actual #203 vs paper #294 91 places worse paper 0.9240 ±0.0030 best at pub: Learning Graph Augmentations to Learn Graph Representations (2022) best today: Learn Beneficial Noise as Graph Augmentation (ICML 2025) paper 0.9240 ±0.0030 Δ at pub 0.0040 Δ today 0.0073
206	MVGRL GNN baseline actual #206 vs paper #206	0.9310 ±0.0040	actual #206 vs paper #206 paper 0.9310 ±0.0040 prev best: Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating (AAAI 2022) paper 0.9310 ±0.0040 Δ at pub 0.0109
213	DGI GNN baseline actual #213 vs paper #340 127 places worse	0.9200 ±0.0060	actual #213 vs paper #340 127 places worse paper 0.9200 ±0.0060 prev best: Graph Contrastive Learning with Adaptive Augmentation (WWW 2020) best today: Noise-Resilient Unsupervised Graph Representation Learning via Multi-Hop Feature Quality Estimation (2024) Improvement within reported uncertainty; not highlighted. paper 0.9200 ±0.0060 Δ at pub 0.0039 Δ today 0.0109
283	PairSim GNN baseline	0.9250 ±0.0060	This paper
301	Par GNN baseline	0.9240 ±0.0050	This paper
312	Clu GNN baseline	0.9230 ±0.0040	This paper
311	GRLC GNN baseline actual #311 vs paper #348 37 places worse	0.9230 ±0.0050 best reported	actual #311 vs paper #348 37 places worse paper 0.9190 ±0.0060 best at pub: Smoothed Graph Contrastive Learning via Seamless Proximity Integration (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.9190 ±0.0060 Δ at pub 0.0040
334	PairDis GNN baseline	0.9210 ±0.0050	This paper
330	CG3 Hyb baseline actual #330 vs paper #330	0.9210 ±0.0050	actual #330 vs paper #330 paper 0.9210 ±0.0050 prev best: Rethinking and Scaling Up Graph Contrastive Learning: An Extremely Efficient Approach with Group Discrimination (NeurIPS 2022) paper 0.9210 ±0.0050 Δ at pub 0.0270
390	MLP code ↗ Trad global	0.9128 ±0.0063	Clarify Confused Nodes via Separated Learning code ↗

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 code ↗ GNN global	0.9412 ±0.0008	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
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 ↗
139	GRACE GNN baseline actual #139 vs paper #184 45 places worse	0.8953 ±0.0035 best reported	actual #139 vs paper #184 45 places worse paper 0.8850 ±0.0050 best at pub: Spectral Augmentation for Self-Supervised Learning on Graphs (ICLR 2022) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) paper 0.8850 ±0.0050 Δ at pub 0.0103 Δ today 0.0103
160	GCA GNN baseline actual #160 vs paper #171 11 places worse	0.8894 ±0.0015 best reported	actual #160 vs paper #171 11 places worse paper 0.8870 ±0.0020 best at pub: Barlow Graph Auto-Encoder for Unsupervised Network Embedding (2021) best today: Adaptive Negative Scheduling for Graph Contrastive Learning (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.8870 ±0.0020 Δ at pub 0.0024 Δ today 0.0024
164	WAS Hyb ★ proposed	0.8890 ±0.0030	This paper
174	AGSSL GNN baseline	0.8870 ±0.0040	This paper
179	MVGRL GNN baseline actual #179 vs paper #212 33 places worse	0.8861 ±0.0064 best reported	actual #179 vs paper #212 33 places worse paper 0.8790 ±0.0020 best at pub: Training MLPs on Graphs without Supervision (2024) best today: Balancing Graph Embedding Smoothness in Self-Supervised Learning via Information-Theoretic Decomposition (WWW 2025) Difference within reported uncertainty; not counted as underperformance. paper 0.8790 ±0.0020 Δ today 0.0071
184	GRLC GNN baseline actual #184 vs paper #206 22 places worse	0.8854 ±0.0023 best reported	actual #184 vs paper #206 22 places worse paper 0.8810 ±0.0040 best at pub: Smoothed Graph Contrastive Learning via Seamless Proximity Integration (2024) Difference within reported uncertainty; not counted as underperformance. paper 0.8810 ±0.0040 Δ at pub 0.0044
188	AutoSSL GNN baseline actual #188 vs paper #188	0.8840 ±0.0030	actual #188 vs paper #188 paper 0.8840 ±0.0030 prev best: Graph Soft-Contrastive Learning via Neighborhood Ranking (2022) Improvement within reported uncertainty; not highlighted. paper 0.8840 ±0.0030 Δ at pub 0.0022
210	DGI GNN baseline actual #210 vs paper #240 30 places worse	0.8740 ±0.0070	actual #210 vs paper #240 30 places worse paper 0.8740 ±0.0070 prev best: Unifying Graph Contrastive Learning with Flexible Contextual Scopes (2022) paper 0.8740 ±0.0070 Δ at pub 0.0330
213	ParetoGNN GNN baseline	0.8790 ±0.0030	This paper
218	CG3 Hyb baseline actual #218 vs paper #218	0.8780 ±0.0090	actual #218 vs paper #218 paper 0.8780 ±0.0090 prev best: Rethinking and Scaling Up Graph Contrastive Learning: An Extremely Efficient Approach with Group Discrimination (NeurIPS 2022) paper 0.8780 ±0.0090 Δ at pub 0.0790
232	Clu GNN baseline	0.8760 ±0.0060	This paper
252	PairSim GNN baseline	0.8710 ±0.0090	This paper
263	Par GNN baseline	0.8670 ±0.0080	This paper
271	PairDis GNN baseline	0.8650 ±0.0070	This paper
320	MLP code ↗ Trad global	0.8463	GraphFM: A generalist graph transformer that learns transferable representations across diverse … code ↗

Coauthor-Physics

Global leaderboard →

#	Model	Accuracy ↑	Paper
1	GrokFormer code ↗ GT global	0.9831 ±0.0018	GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers code ↗
2	PolyFormer code ↗ GT global	0.9808 ±0.0027	GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers code ↗
3	SGFormer code ↗ GT global	0.9796 ±0.0081	GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers code ↗
9	GCN code ↗ GNN global	0.9774 ±0.0035	GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers code ↗
58	MLP code ↗ Trad global	0.9597 ±0.0007	Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification code ↗
63	GCA GNN baseline actual #63 vs paper #192 129 places worse	0.9590 ±0.0020 best reported	actual #63 vs paper #192 129 places worse paper 0.9410 ±0.0050 best at pub: Unifying Graph Contrastive Learning with Flexible Contextual Scopes (2022) paper 0.9410 ±0.0050 Δ at pub 0.0180
77	GRACE GNN baseline actual #77 vs paper #174 97 places worse	0.9574 ±0.0006 best reported	actual #77 vs paper #174 97 places worse paper 0.9450 ±0.0040 best at pub: Building Shortcuts between Distant Nodes with Biaffine Mapping for Graph Convolutional Networks (2023) best today: Learn Beneficial Noise as Graph Augmentation (ICML 2025) paper 0.9450 ±0.0040 Δ at pub 0.0113 Δ today 0.0124
166	WAS Hyb ★ proposed	0.9490 ±0.0030	This paper
167	DGI GNN baseline actual #167 vs paper #201 34 places worse	0.9489 ±0.0009 best reported	actual #167 vs paper #201 34 places worse paper 0.9390 ±0.0060 best at pub: Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating (AAAI 2022) best today: Balancing Graph Embedding Smoothness in Self-Supervised Learning via Information-Theoretic Decomposition (WWW 2025) paper 0.9390 ±0.0060 Δ at pub 0.0067 Δ today 0.0099
168	AGSSL GNN baseline	0.9480 ±0.0030	This paper
180	ParetoGNN GNN baseline	0.9440 ±0.0060	This paper
200	PairDis GNN baseline	0.9400 ±0.0060	This paper
204	AutoSSL GNN baseline	0.9390 ±0.0060	This paper
209	CG3 Hyb baseline	0.9360 ±0.0060	This paper
218	Clu GNN baseline	0.9350 ±0.0050	This paper
224	PairSim GNN baseline	0.9330 ±0.0030	This paper
231	GRLC GNN baseline	0.9320 ±0.0050	This paper
238	Par GNN baseline	0.9310 ±0.0080	This paper

MoleculeNet

Tox21

Global leaderboard →

#	Model	ROC-AUC ↑	Paper
1	HIMP Hyb global	0.8740 ±0.0050	Graph convolutions that can finally model local structure
2	GINE w/ VN GNN global	0.8720 ±0.0030	Graph convolutions that can finally model local structure
3	NaiveGINe+ K=3 w/ VN GNN global	0.8700 ±0.0040	Graph convolutions that can finally model local structure
14	GCN GNN global	0.8400 ±0.0040	Graph convolutions that can finally model local structure
31	JOAO GNN baseline actual #31 vs paper #238 207 places worse	0.8230 ±0.0220 best reported	actual #31 vs paper #238 207 places worse paper 0.7480 ±0.0060 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.7480 ±0.0060 Δ at pub 0.0750
35	EdgePred GNN baseline actual #35 vs paper #312 277 places worse	0.8180 ±0.0250 best reported	actual #35 vs paper #312 277 places worse paper 0.6820 ±0.0910 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.6820 ±0.0910 Δ at pub 0.1360
46	ContextPred GNN baseline actual #46 vs paper #283 237 places worse	0.8060 ±0.0120 best reported	actual #46 vs paper #283 237 places worse paper 0.7290 ±0.0080 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.7290 ±0.0080 Δ at pub 0.0770
47	GraphCL GNN baseline actual #47 vs paper #228 181 places worse	0.8050 ±0.0170 best reported	actual #47 vs paper #228 181 places worse paper 0.7510 ±0.0080 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.7510 ±0.0080 Δ at pub 0.0540
61	GraphMVP GNN baseline actual #61 vs paper #242 181 places worse	0.7990 ±0.0180 best reported	actual #61 vs paper #242 181 places worse paper 0.7460 ±0.0030 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.7460 ±0.0030 Δ at pub 0.0530
71	GraphLoG GNN baseline actual #71 vs paper #265 194 places worse	0.7960 ±0.0250 best reported	actual #71 vs paper #265 194 places worse paper 0.7390 ±0.0140 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.7390 ±0.0140 Δ at pub 0.0570
81	InfoGraph GNN baseline actual #81 vs paper #263 182 places worse	0.7930 best reported	actual #81 vs paper #263 182 places worse paper 0.7400 ±0.0050 best at pub: GeomGCL: Geometric Graph Contrastive Learning for Molecular Property Prediction (AAAI 2021) paper 0.7400 ±0.0050 Δ at pub 0.0530
180	GPT-GNN LLM baseline actual #180 vs paper #253 73 places worse	0.7610 ±0.0040 best reported	actual #180 vs paper #253 73 places worse paper 0.7430 ±0.0070 best at pub: Motif-based Graph Self-Supervised Learning for Molecular Property Prediction (NeurIPS 2021) paper 0.7430 ±0.0070 Δ at pub 0.0180
202	GraphMAE GNN baseline actual #202 vs paper #310 108 places worse	0.7550 ±0.0060 best reported	actual #202 vs paper #310 108 places worse paper 0.6880 ±0.0050 best at pub: GraphMAE: Self-Supervised Masked Graph Autoencoders (KDD 2022) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.6880 ±0.0050 Δ at pub 0.0670 Δ today 0.0670
209	WAS Hyb ★ proposed	0.7540 ±0.0070	This paper
223	R-GCL GNN baseline	0.7520 ±0.0030	This paper
220	G-Contextual GNN baseline actual #220 vs paper #235 15 places worse	0.7520 ±0.0030 best reported	actual #220 vs paper #235 15 places worse paper 0.7490 ±0.0020 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.7490 ±0.0020 Δ at pub 0.0030
229	AttrMask GNN baseline actual #229 vs paper #247 18 places worse	0.7510 ±0.0090 best reported	actual #229 vs paper #247 18 places worse paper 0.7440 ±0.0050 best at pub: Motif-aware Attribute Masking for Molecular Graph Pre-training (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.7440 ±0.0050 Δ at pub 0.0070
240	AD-GCL GNN baseline actual #240 vs paper #240	0.7470 ±0.0110	actual #240 vs paper #240 paper 0.7470 ±0.0110 prev best: Graph Self-Contrast Representation Learning (2023) Improvement within reported uncertainty; not highlighted. paper 0.7470 ±0.0110 Δ at pub 0.0114
257	Random-Select GNN baseline	0.7430 ±0.0060	This paper
256	SimGRACE GNN baseline actual #256 vs paper #256	0.7430 ±0.0020	actual #256 vs paper #256 paper 0.7430 ±0.0020 prev best: Graph Self-supervised Learning with Accurate Discrepancy Learning (NeurIPS 2022) Improvement within reported uncertainty; not highlighted. paper 0.7430 ±0.0020 Δ at pub 0.0010
280	G-Motif GNN baseline actual #280 vs paper #281 1 place worse	0.7320 ±0.0080 best reported	actual #280 vs paper #281 1 place worse paper 0.7310 ±0.0040 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.7310 ±0.0040 Δ at pub 0.0010
314	Average-Weight GNN baseline	0.6780 ±0.0080	This paper

BACE

Global leaderboard →

#	Model	ROC-AUC ↑	Paper
1	ProtoMol Hyb global	0.9140 ±0.0030	ProtoMol: Enhancing Molecular Property Prediction via Prototype-Guided Multimodal Learning
2	MMSG GNN global	0.9080 ±0.0050	ProtoMol: Enhancing Molecular Property Prediction via Prototype-Guided Multimodal Learning
3	MemGNN code ↗ Hyb global	0.9070	Memory-Based Graph Networks code ↗
43	GIN code ↗ GNN global	0.8646 ±0.0081	Pre-training Graph Neural Networks on Molecules by Using Subgraph-Conditioned Graph Information … code ↗
98	GraphLoG GNN baseline actual #98 vs paper #214 116 places worse	0.8350 ±0.0120 best reported	actual #98 vs paper #214 116 places worse paper 0.7810 ±0.0100 best at pub: Motif-aware Attribute Masking for Molecular Graph Pre-training (2023) best today: UGMAE: A Unified Framework for Graph Masked Autoencoders (2024) paper 0.7810 ±0.0100 Δ at pub 0.0540 Δ today 0.0540
100	JOAO GNN baseline actual #100 vs paper #257 157 places worse	0.8340 ±0.0330 best reported	actual #100 vs paper #257 157 places worse paper 0.7340 ±0.0120 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.7340 ±0.0120 Δ at pub 0.1000
105	GraphMAE GNN baseline actual #105 vs paper #250 145 places worse	0.8310 ±0.0090 best reported	actual #105 vs paper #250 145 places worse paper 0.7500 ±0.0070 best at pub: GraphMAE: Self-Supervised Masked Graph Autoencoders (KDD 2022) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.7500 ±0.0070 Δ at pub 0.0810 Δ today 0.0810
149	GraphMVP GNN baseline actual #149 vs paper #171 22 places worse	0.8120 ±0.0090 best reported	actual #149 vs paper #171 22 places worse paper 0.8030 ±0.0140 best at pub: Unified 2D and 3D Pre-Training of Molecular Representations (KDD 2022) best today: Invariant-Based Diagnostics for Graph Benchmarks (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.8030 ±0.0140 Δ at pub 0.0090 Δ today 0.0090
165	WAS Hyb ★ proposed	0.8070 ±0.0050	This paper
176	EdgePred GNN baseline actual #176 vs paper #292 116 places worse	0.7990 ±0.0090 best reported	actual #176 vs paper #292 116 places worse paper 0.6610 ±0.0260 best at pub: Graph Self-Contrast Representation Learning (2023) paper 0.6610 ±0.0260 Δ at pub 0.1380
182	AttrMask GNN baseline actual #182 vs paper #226 44 places worse	0.7970 ±0.0033 best reported	actual #182 vs paper #226 44 places worse paper 0.7740 ±0.0020 best at pub: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining (ICML 2023) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.7740 ±0.0020 Δ at pub 0.0230 Δ today 0.0230
186	ContextPred GNN baseline actual #186 vs paper #226 40 places worse	0.7960 ±0.0120 best reported	actual #186 vs paper #226 40 places worse paper 0.7730 ±0.0100 best at pub: Graph Self-Contrast Representation Learning (2023) best today: Dual-perspective Cross Contrastive Learning in Graph Transformers (2024) paper 0.7730 ±0.0100 Δ at pub 0.0230 Δ today 0.0230
196	G-Contextual GNN baseline actual #196 vs paper #240 44 places worse	0.7920 ±0.0030 best reported	actual #196 vs paper #240 44 places worse paper 0.7630 ±0.0190 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.7630 ±0.0190 Δ at pub 0.0290
202	AD-GCL GNN baseline actual #202 vs paper #240 38 places worse	0.7630 ±0.0060	actual #202 vs paper #240 38 places worse paper 0.7630 ±0.0060 prev best: Let Invariant Rationale Discovery Inspire Graph Contrastive Learning (ICML 2022) best today: Community-Invariant Graph Contrastive Learning (ICML 2024) Improvement within reported uncertainty; not highlighted. paper 0.7630 ±0.0060 Δ at pub 0.0128 Δ today 0.0270
204	InfoGraph GNN baseline actual #204 vs paper #273 69 places worse	0.7890 ±0.0110 best reported	actual #204 vs paper #273 69 places worse paper 0.7140 ±0.0240 best at pub: Omni-Granular Ego-Semantic Propagation for Self-Supervised Graph Representation Learning (ICML 2022) paper 0.7140 ±0.0240 Δ at pub 0.0750
208	GPT-GNN LLM baseline actual #208 vs paper #208	0.7860 ±0.0290	actual #208 vs paper #208 paper 0.7860 ±0.0290 prev best: Motif-based Graph Self-Supervised Learning for Molecular Property Prediction (NeurIPS 2021) Improvement within reported uncertainty; not highlighted. paper 0.7860 ±0.0290 Δ at pub 0.0010
213	GraphCL GNN baseline actual #213 vs paper #226 13 places worse	0.7820 ±0.0120 best reported	actual #213 vs paper #226 13 places worse paper 0.7750 ±0.0160 best at pub: Omni-Granular Ego-Semantic Propagation for Self-Supervised Graph Representation Learning (ICML 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.7750 ±0.0160 Δ at pub 0.0070
227	Random-Select GNN baseline	0.7720 ±0.0170	This paper
244	SimGRACE GNN baseline actual #244 vs paper #244	0.7580 ±0.0040	actual #244 vs paper #244 paper 0.7580 ±0.0040 prev best: Graph Self-supervised Learning with Accurate Discrepancy Learning (NeurIPS 2022) Improvement within reported uncertainty; not highlighted. paper 0.7580 ±0.0040 Δ at pub 0.0199
249	Average-Weight GNN baseline	0.7530 ±0.0120	This paper
256	R-GCL GNN baseline	0.7360 ±0.0120	This paper
259	G-Motif GNN baseline actual #259 vs paper #270 11 places worse	0.7340 ±0.0400 best reported	actual #259 vs paper #270 11 places worse paper 0.7150 ±0.0090 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.7150 ±0.0090 Δ at pub 0.0190

BBBP

Global leaderboard →

#	Model	ROC-AUC ↑	Paper
1	MAGPrompt+ Hyb global	0.9432 ±0.0238	MAGPrompt: Message-Adaptive Graph Prompt Tuning for Graph Neural Networks
2	CamS-LLaMA LLM 100.0M global	0.9420	Multi-scale Graph Autoregressive Modeling: Molecular Property Prediction via Next Token …
3	SE(3)-I MPPH Hyb global	0.9400 ±0.0210	SE(3)-Invariant Multiparameter Persistent Homology for Chiral-Sensitive Molecular Property …
34	GIN code ↗ GNN global	0.8875 ±0.0049	Pre-training Graph Neural Networks on Molecules by Using Subgraph-Conditioned Graph Information … code ↗
36	GraphCL GNN baseline actual #36 vs paper #154 118 places worse	0.8870 ±0.0190 best reported	actual #36 vs paper #154 118 places worse paper 0.6990 ±0.0160 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6990 ±0.0160 Δ at pub 0.1880
39	JOAO GNN baseline actual #39 vs paper #275 236 places worse	0.8790 ±0.0200 best reported	actual #39 vs paper #275 236 places worse paper 0.6470 ±0.0080 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6470 ±0.0080 Δ at pub 0.2320
40	ContextPred GNN baseline actual #40 vs paper #179 139 places worse	0.8770 ±0.0260 best reported	actual #40 vs paper #179 139 places worse paper 0.6900 ±0.0200 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6900 ±0.0200 Δ at pub 0.1870
44	EdgePred GNN baseline actual #44 vs paper #188 144 places worse	0.8730 ±0.0160 best reported	actual #44 vs paper #188 144 places worse paper 0.6860 ±0.0670 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6860 ±0.0670 Δ at pub 0.1870
49	GraphMVP GNN baseline actual #49 vs paper #201 152 places worse	0.8600 ±0.0340 best reported	actual #49 vs paper #201 152 places worse paper 0.6790 ±0.0100 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6790 ±0.0100 Δ at pub 0.1810
54	GraphLoG GNN baseline actual #54 vs paper #245 191 places worse	0.8460 ±0.0080 best reported	actual #54 vs paper #245 191 places worse paper 0.6640 ±0.0280 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6640 ±0.0280 Δ at pub 0.1820
101	GraphMAE GNN baseline actual #101 vs paper #248 147 places worse	0.7200 ±0.0060 best reported	actual #101 vs paper #248 147 places worse paper 0.6630 ±0.0070 best at pub: GraphMAE: Self-Supervised Masked Graph Autoencoders (KDD 2022) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.6630 ±0.0070 Δ at pub 0.0570 Δ today 0.0570
118	SimGRACE GNN baseline actual #118 vs paper #240 122 places worse	0.7125 ±0.0086 best reported	actual #118 vs paper #240 122 places worse paper 0.6670 ±0.0060 best at pub: Graph Self-supervised Learning with Accurate Discrepancy Learning (NeurIPS 2022) best today: Graph Self-Contrast Representation Learning (2023) paper 0.6670 ±0.0060 Δ at pub 0.0455 Δ today 0.0455
130	WAS Hyb ★ proposed	0.7050 ±0.0100	This paper
135	G-Contextual GNN baseline actual #135 vs paper #282 147 places worse	0.7030 ±0.0160 best reported	actual #135 vs paper #282 147 places worse paper 0.6400 ±0.0150 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) paper 0.6400 ±0.0150 Δ at pub 0.0630
136	G-Motif GNN baseline actual #136 vs paper #136	0.7030 ±0.0150	actual #136 vs paper #136 paper 0.7030 ±0.0150 prev best: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Improvement within reported uncertainty; not highlighted. paper 0.7030 ±0.0150 Δ at pub 0.0390
141	AttrMask GNN baseline actual #141 vs paper #268 127 places worse	0.7020 ±0.0050 best reported	actual #141 vs paper #268 127 places worse paper 0.6530 ±0.0160 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) best today: Graph Generative Pre-trained Transformer (ICML 2025) paper 0.6530 ±0.0160 Δ at pub 0.0490 Δ today 0.0490
144	AD-GCL GNN baseline actual #144 vs paper #144	0.7010 ±0.0120	actual #144 vs paper #144 paper 0.7010 ±0.0120 prev best: Graph Self-Contrast Representation Learning (2023) Improvement within reported uncertainty; not highlighted. paper 0.7010 ±0.0120 Δ at pub 0.0184
170	R-GCL GNN baseline	0.6930 ±0.0040	This paper
173	InfoGraph GNN baseline actual #173 vs paper #250 77 places worse	0.6920 ±0.0080 best reported	actual #173 vs paper #250 77 places worse paper 0.6620 ±0.0110 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) best today: Graph Generative Pre-trained Transformer (ICML 2025) paper 0.6620 ±0.0110 Δ at pub 0.0300 Δ today 0.0300
220	GPT-GNN LLM baseline actual #220 vs paper #268 48 places worse	0.6750 ±0.0130 best reported	actual #220 vs paper #268 48 places worse paper 0.6530 ±0.0150 best at pub: Motif-based Graph Self-Supervised Learning for Molecular Property Prediction (NeurIPS 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.6530 ±0.0150 Δ at pub 0.0220
227	Random-Select GNN baseline	0.6720 ±0.0050	This paper
242	Average-Weight GNN baseline	0.6660 ±0.0020	This paper

ClinTox

Global leaderboard →

#	Model	ROC-AUC ↑	Paper
1	SE(3)-I MPPH Hyb global	0.9930 ±0.0040	SE(3)-Invariant Multiparameter Persistent Homology for Chiral-Sensitive Molecular Property …
2	KA-GCN code ↗ GNN global	0.9920 ±0.0050	A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the … code ↗
3	KA-GAT code ↗ GNN global	0.9910 ±0.0050	A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the … code ↗
19	GCN GNN global	0.9130 ±0.0170	OOD-GNN: Out-of-Distribution Generalized Graph Neural Network
40	AttrMask GNN baseline actual #40 vs paper #171 131 places worse	0.8770 ±0.0119 best reported	actual #40 vs paper #171 131 places worse paper 0.7030 ±0.0750 best at pub: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining (ICML 2023) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.7030 ±0.0750 Δ at pub 0.1740 Δ today 0.1740
49	GraphMVP GNN baseline actual #49 vs paper #118 69 places worse	0.8650 ±0.0170 best reported	actual #49 vs paper #118 69 places worse paper 0.7640 ±0.0110 best at pub: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining (ICML 2023) best today: SubGDiff: A Subgraph Diffusion Model to Improve Molecular Representation Learning (NeurIPS 2024) paper 0.7640 ±0.0110 Δ at pub 0.1010 Δ today 0.1010
71	GraphMAE GNN baseline actual #71 vs paper #200 129 places worse	0.8230 ±0.0120 best reported	actual #71 vs paper #200 129 places worse paper 0.6530 ±0.0390 best at pub: GraphMAE: Self-Supervised Masked Graph Autoencoders (KDD 2022) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.6530 ±0.0390 Δ at pub 0.1700 Δ today 0.1700
78	G-Motif GNN baseline actual #78 vs paper #78	0.8140 ±0.0220	actual #78 vs paper #78 paper 0.8140 ±0.0220 prev best: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Improvement within reported uncertainty; not highlighted. paper 0.8140 ±0.0220 Δ at pub 0.0360
79	JOAO GNN baseline actual #79 vs paper #195 116 places worse	0.8132 ±0.0249 best reported	actual #79 vs paper #195 116 places worse paper 0.6610 ±0.0370 best at pub: Graph Self-Contrast Representation Learning (2023) paper 0.6610 ±0.0370 Δ at pub 0.1522
82	AD-GCL GNN baseline actual #82 vs paper #105 23 places worse	0.7810 ±0.0360	actual #82 vs paper #105 23 places worse paper 0.7810 ±0.0360 prev best: Let Invariant Rationale Discovery Inspire Graph Contrastive Learning (ICML 2022) Improvement within reported uncertainty; not highlighted. paper 0.7810 ±0.0360 Δ at pub 0.0047
97	GraphCL GNN baseline actual #97 vs paper #158 61 places worse	0.7890 ±0.0420 best reported	actual #97 vs paper #158 61 places worse paper 0.7210 ±0.0470 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) best today: Improving Molecular Pretraining with Complementary Featurizations (2022) Difference within reported uncertainty; not counted as underperformance. paper 0.7210 ±0.0470 Δ at pub 0.0680 Δ today 0.0680
99	R-GCL GNN baseline	0.7890 ±0.0410	This paper
103	GraphLoG GNN baseline actual #103 vs paper #207 104 places worse	0.7872 ±0.0258 best reported	actual #103 vs paper #207 104 places worse paper 0.6410 ±0.0340 best at pub: Let Invariant Rationale Discovery Inspire Graph Contrastive Learning (ICML 2022) best today: Graph Self-Contrast Representation Learning (2023) paper 0.6410 ±0.0340 Δ at pub 0.1462 Δ today 0.1462
106	InfoGraph GNN baseline actual #106 vs paper #164 58 places worse	0.7810 best reported	actual #106 vs paper #164 58 places worse paper 0.7150 ±0.0630 best at pub: GeomGCL: Geometric Graph Contrastive Learning for Molecular Property Prediction (AAAI 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.7150 ±0.0630 Δ at pub 0.0660
115	ContextPred GNN baseline actual #115 vs paper #194 79 places worse	0.7720 ±0.0088 best reported	actual #115 vs paper #194 79 places worse paper 0.6690 ±0.0760 best at pub: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining (ICML 2023) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) Difference within reported uncertainty; not counted as underperformance. paper 0.6690 ±0.0760 Δ at pub 0.1030 Δ today 0.1030
123	SimGRACE GNN baseline actual #123 vs paper #194 71 places worse	0.6680 ±0.0200	actual #123 vs paper #194 71 places worse paper 0.6680 ±0.0200 prev best: Graph Self-supervised Learning with Accurate Discrepancy Learning (NeurIPS 2022) best today: Dual-perspective Cross Contrastive Learning in Graph Transformers (2024) Improvement within reported uncertainty; not highlighted. paper 0.6680 ±0.0200 Δ at pub 0.0264 Δ today 0.0880
131	GPT-GNN LLM baseline actual #131 vs paper #235 104 places worse	0.7490 ±0.0270 best reported	actual #131 vs paper #235 104 places worse paper 0.5610 ±0.0890 best at pub: Motif-based Graph Self-Supervised Learning for Molecular Property Prediction (NeurIPS 2021) paper 0.5610 ±0.0890 Δ at pub 0.1880
150	WAS Hyb ★ proposed	0.7310 ±0.0050	This paper
191	G-Contextual GNN baseline actual #191 vs paper #191	0.6760 ±0.0290	actual #191 vs paper #191 paper 0.6760 ±0.0290 prev best: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Improvement within reported uncertainty; not highlighted. paper 0.6760 ±0.0290 Δ at pub 0.0770
192	EdgePred GNN baseline actual #192 vs paper #192	0.6730 ±0.0200	actual #192 vs paper #192 paper 0.6730 ±0.0200 prev best: Subgraph Networks Based Contrastive Learning (2023) Improvement within reported uncertainty; not highlighted. paper 0.6730 ±0.0200 Δ at pub 0.0320
216	Average-Weight GNN baseline	0.6260 ±0.0440	This paper
230	Random-Select GNN baseline	0.5810 ±0.0150	This paper

SIDER

Global leaderboard →

#	Model	ROC-AUC ↑	Paper
1	KA-GAT code ↗ GNN global	0.8470 ±0.0020	A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the … code ↗
2	KA-GCN code ↗ GNN global	0.8420 ±0.0010	A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the … code ↗
3	GraphKAN code ↗ GNN global	0.8370 ±0.0010	A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the … code ↗
66	GIN code ↗ GNN global	0.6403 ±0.0104	Pre-training Graph Neural Networks on Molecules by Using Subgraph-Conditioned Graph Information … code ↗
68	JOAO GNN baseline actual #68 vs paper #150 82 places worse	0.6400 ±0.0100 best reported	actual #68 vs paper #150 82 places worse paper 0.6070 ±0.0100 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6070 ±0.0100 Δ at pub 0.0330
70	GraphMVP GNN baseline actual #70 vs paper #176 106 places worse	0.6390 ±0.0120 best reported	actual #70 vs paper #176 106 places worse paper 0.6020 ±0.0160 best at pub: Unified 2D and 3D Pre-Training of Molecular Representations (KDD 2022) best today: A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the Foundation Model Era (2026) paper 0.6020 ±0.0160 Δ at pub 0.0370 Δ today 0.0370
103	ContextPred GNN baseline actual #103 vs paper #206 103 places worse	0.6220 ±0.0059 best reported	actual #103 vs paper #206 103 places worse paper 0.5870 ±0.0160 best at pub: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining (ICML 2023) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.5870 ±0.0160 Δ at pub 0.0350 Δ today 0.0350
114	GraphCL GNN baseline actual #114 vs paper #183 69 places worse	0.6183 ±0.0060 best reported	actual #114 vs paper #183 69 places worse paper 0.5990 ±0.0150 best at pub: Omni-Granular Ego-Semantic Propagation for Self-Supervised Graph Representation Learning (ICML 2022) best today: Pre-training Graph Neural Networks on Molecules by Using Subgraph-Conditioned Graph Information Bottleneck (AAAI 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.5990 ±0.0150 Δ at pub 0.0090 Δ today 0.0193
123	GraphLoG GNN baseline actual #123 vs paper #195 72 places worse	0.6150 ±0.0130 best reported	actual #123 vs paper #195 72 places worse paper 0.5950 ±0.0240 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) Difference within reported uncertainty; not counted as underperformance. paper 0.5950 ±0.0240 Δ at pub 0.0200
134	AttrMask GNN baseline actual #134 vs paper #232 98 places worse	0.6120 ±0.0012 best reported	actual #134 vs paper #232 98 places worse paper 0.5510 ±0.0070 best at pub: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining (ICML 2023) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.5510 ±0.0070 Δ at pub 0.0610 Δ today 0.0610
139	G-Contextual GNN baseline actual #139 vs paper #139	0.6110 ±0.0250	actual #139 vs paper #139 paper 0.6110 ±0.0250 prev best: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Improvement within reported uncertainty; not highlighted. paper 0.6110 ±0.0250 Δ at pub 0.0270
146	AD-GCL GNN baseline actual #146 vs paper #195 49 places worse	0.5940 ±0.0070	actual #146 vs paper #195 49 places worse paper 0.5940 ±0.0070 prev best: Let Invariant Rationale Discovery Inspire Graph Contrastive Learning (ICML 2022) best today: Community-Invariant Graph Contrastive Learning (ICML 2024) Improvement within reported uncertainty; not highlighted. paper 0.5940 ±0.0070 Δ at pub 0.0016 Δ today 0.0140
158	G-Motif GNN baseline actual #158 vs paper #178 20 places worse	0.6060 ±0.0110 best reported	actual #158 vs paper #178 20 places worse paper 0.6010 ±0.0170 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.6010 ±0.0170 Δ at pub 0.0050
156	SimGRACE GNN baseline actual #156 vs paper #206 50 places worse	0.6060 ±0.0100 best reported	actual #156 vs paper #206 50 places worse paper 0.5840 ±0.0110 best at pub: Graph Self-supervised Learning with Accurate Discrepancy Learning (NeurIPS 2022) best today: Dual-perspective Cross Contrastive Learning in Graph Transformers (2024) paper 0.5840 ±0.0110 Δ at pub 0.0219 Δ today 0.0220
162	InfoGraph GNN baseline actual #162 vs paper #212 50 places worse	0.6054 ±0.0090 best reported	actual #162 vs paper #212 50 places worse paper 0.5820 ±0.0040 best at pub: A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining (ICML 2023) best today: Community-Invariant Graph Contrastive Learning (ICML 2024) paper 0.5820 ±0.0040 Δ at pub 0.0170 Δ today 0.0234
171	WAS Hyb ★ proposed	0.6040 ±0.0040	This paper
170	EdgePred GNN baseline actual #170 vs paper #204 34 places worse	0.6040 ±0.0070 best reported	actual #170 vs paper #204 34 places worse paper 0.5890 ±0.0130 best at pub: Subgraph Networks Based Contrastive Learning (2023) best today: Community-Invariant Graph Contrastive Learning (ICML 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.5890 ±0.0130 Δ at pub 0.0150 Δ today 0.0150
175	GraphMAE GNN baseline actual #175 vs paper #223 48 places worse	0.6030 ±0.0110 best reported	actual #175 vs paper #223 48 places worse paper 0.5710 ±0.0070 best at pub: GraphMAE: Self-Supervised Masked Graph Autoencoders (KDD 2022) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.5710 ±0.0070 Δ at pub 0.0320 Δ today 0.0320
184	R-GCL GNN baseline	0.5990 ±0.0090	This paper
197	GPT-GNN LLM baseline actual #197 vs paper #215 18 places worse	0.5930 ±0.0080 best reported	actual #197 vs paper #215 18 places worse paper 0.5790 ±0.0020 best at pub: Motif-based Graph Self-Supervised Learning for Molecular Property Prediction (NeurIPS 2021) paper 0.5790 ±0.0020 Δ at pub 0.0140
200	Random-Select GNN baseline	0.5920 ±0.0170	This paper
225	Average-Weight GNN baseline	0.5690 ±0.0150	This paper

ToxCast

Global leaderboard →

#	Model	ROC-AUC ↑	Paper
1	GeomGCL GNN global	0.7630	GeomGCL: Geometric Graph Contrastive Learning for Molecular Property Prediction
2	LiGhT DeepGPT ext data Hyb 0.4M global	0.7570 ±0.0110	Deep Prompt Tuning for Graph Transformers
3	CoMPT GT global	0.7550	GeomGCL: Geometric Graph Contrastive Learning for Molecular Property Prediction
14	GCN GNN global	0.7350 ±0.0020	Graph convolutions that can finally model local structure
28	ContextPred GNN baseline actual #28 vs paper #185 157 places worse	0.7140 ±0.0180 best reported	actual #28 vs paper #185 157 places worse paper 0.6170 ±0.0070 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6170 ±0.0070 Δ at pub 0.0970
30	EdgePred GNN baseline actual #30 vs paper #204 174 places worse	0.7120 ±0.0110 best reported	actual #30 vs paper #204 174 places worse paper 0.5900 ±0.0090 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.5900 ±0.0090 Δ at pub 0.1220
31	JOAO GNN baseline actual #31 vs paper #178 147 places worse	0.7110 ±0.0140 best reported	actual #31 vs paper #178 147 places worse paper 0.6240 ±0.0040 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6240 ±0.0040 Δ at pub 0.0870
35	InfoGraph GNN baseline actual #35 vs paper #196 161 places worse	0.7050 best reported	actual #35 vs paper #196 161 places worse paper 0.6070 ±0.0090 best at pub: GeomGCL: Geometric Graph Contrastive Learning for Molecular Property Prediction (AAAI 2021) paper 0.6070 ±0.0090 Δ at pub 0.0980
42	GraphCL GNN baseline actual #42 vs paper #167 125 places worse	0.6960 ±0.0230 best reported	actual #42 vs paper #167 125 places worse paper 0.6280 ±0.0070 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6280 ±0.0070 Δ at pub 0.0680
50	GraphMVP GNN baseline actual #50 vs paper #137 87 places worse	0.6890 ±0.0100 best reported	actual #50 vs paper #137 87 places worse paper 0.6350 ±0.0020 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6350 ±0.0020 Δ at pub 0.0540
64	GraphLoG GNN baseline actual #64 vs paper #180 116 places worse	0.6770 ±0.0190 best reported	actual #64 vs paper #180 116 places worse paper 0.6230 ±0.0060 best at pub: KPGT: Knowledge-Guided Pre-training of Graph Transformer for Molecular Property Prediction (KDD 2022) paper 0.6230 ±0.0060 Δ at pub 0.0540
128	GraphMAE GNN baseline actual #128 vs paper #186 58 places worse	0.6410 ±0.0030 best reported	actual #128 vs paper #186 58 places worse paper 0.6150 ±0.0050 best at pub: GraphMAE: Self-Supervised Masked Graph Autoencoders (KDD 2022) best today: A Generalist Cross-Domain Molecular Learning Framework for Structure-Based Drug Discovery (2025) paper 0.6150 ±0.0050 Δ at pub 0.0260 Δ today 0.0260
136	WAS Hyb ★ proposed	0.6370 ±0.0150	This paper
138	AD-GCL GNN baseline actual #138 vs paper #167 29 places worse	0.6350 ±0.0070 best reported	actual #138 vs paper #167 29 places worse paper 0.6280 ±0.0050 best at pub: Let Invariant Rationale Discovery Inspire Graph Contrastive Learning (ICML 2022) best today: Community-Invariant Graph Contrastive Learning (ICML 2024) Difference within reported uncertainty; not counted as underperformance. paper 0.6280 ±0.0050 Δ at pub 0.0030 Δ today 0.0070
145	SimGRACE GNN baseline actual #145 vs paper #182 37 places worse	0.6336 ±0.0052 best reported	actual #145 vs paper #182 37 places worse paper 0.6210 ±0.0010 best at pub: Graph Self-supervised Learning with Accurate Discrepancy Learning (NeurIPS 2022) best today: Graph Self-Contrast Representation Learning (2023) paper 0.6210 ±0.0010 Δ at pub 0.0126 Δ today 0.0126
150	AttrMask GNN baseline actual #150 vs paper #171 21 places worse	0.6330 ±0.0060 best reported	actual #150 vs paper #171 21 places worse paper 0.6260 ±0.0010 best at pub: Motif-aware Attribute Masking for Molecular Graph Pre-training (2023) Difference within reported uncertainty; not counted as underperformance. paper 0.6260 ±0.0010 Δ at pub 0.0070
151	GPT-GNN LLM baseline actual #151 vs paper #151	0.6330 ±0.0030	actual #151 vs paper #151 paper 0.6330 ±0.0030 prev best: Motif-based Graph Self-Supervised Learning for Molecular Property Prediction (NeurIPS 2021) Improvement within reported uncertainty; not highlighted. paper 0.6330 ±0.0030 Δ at pub 0.0020
172	G-Contextual GNN baseline actual #172 vs paper #185 13 places worse	0.6260 ±0.0030 best reported	actual #172 vs paper #185 13 places worse paper 0.6170 ±0.0010 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) paper 0.6170 ±0.0010 Δ at pub 0.0090
173	G-Motif GNN baseline actual #173 vs paper #185 12 places worse	0.6260 ±0.0050 best reported	actual #173 vs paper #185 12 places worse paper 0.6180 ±0.0060 best at pub: 16 Pre-training Molecular Graph Representation with 3D Geometry (ICLR 2021) Difference within reported uncertainty; not counted as underperformance. paper 0.6180 ±0.0060 Δ at pub 0.0080
180	R-GCL GNN baseline	0.6240 ±0.0060	This paper
182	Average-Weight GNN baseline	0.6220 ±0.0180	This paper
203	Random-Select GNN baseline	0.5930 ±0.0070	This paper

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