Classic
ZINC
Structure
249,456Graphs
23Nodes
25Edges
NoneClasses
28Node Feature Dim
4Edge Feature Dim
Coverage
137Results
1Variants
28Papers
- FloydNet: A Learning Paradigm for Global Relational Reasoning FloydNet: A Learning Paradigm for Global Relational Reasoning
- A Hierarchical Quantized Tokenization Framework for Task-Adaptive Graph Representation Learning A Hierarchical Quantized Tokenization Framework for Task-Adaptive Graph Representation Learning
- A Graph Laplacian Eigenvector-based Pre-training Method for Graph Neural Networks A Graph Laplacian Eigenvector-based Pre-training Method for Graph Neural Networks
- ICML 2025: Positional Encoding meets Persistent Homology on Graphs Positional Encoding meets Persistent Homology on Graphs ICML 2025
- Using Random Noise Equivariantly to Boost Graph Neural Networks Universally Using Random Noise Equivariantly to Boost Graph Neural Networks Universally
- ICML 2025: Beyond Message Passing: Neural Graph Pattern Machine Beyond Message Passing: Neural Graph Pattern Machine ICML 2025
- Graph in Graph Neural Network Graph in Graph Neural Network
- NeurIPS 2024: A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening NeurIPS 2024
- What Can We Learn from State Space Models for Machine Learning on Graphs? What Can We Learn from State Space Models for Machine Learning on Graphs?
- ICML 2024: Graph as Point Set Graph as Point Set ICML 2024
- ICML 2024: Subgraphormer: Unifying Subgraph GNNs and Graph Transformers via Graph Products Subgraphormer: Unifying Subgraph GNNs and Graph Transformers via Graph Products ICML 2024
- Topology-Informed Graph Transformer Topology-Informed Graph Transformer
- NeurIPS 2024: Towards Principled Graph Transformers Towards Principled Graph Transformers NeurIPS 2024
- Enhancing Graph Neural Networks with Quantum Computed Encodings Enhancing Graph Neural Networks with Quantum Computed Encodings
- NeurIPS 2023: MAG-GNN: Reinforcement Learning Boosted Graph Neural Network MAG-GNN: Reinforcement Learning Boosted Graph Neural Network NeurIPS 2023
- NeurIPS 2023: Distance-Restricted Folklore Weisfeiler-Leman GNNs with Provable Cycle Counting Power Distance-Restricted Folklore Weisfeiler-Leman GNNs with Provable Cycle Counting Power NeurIPS 2023
- The Expressive Power of Graph Neural Networks: A Survey The Expressive Power of Graph Neural Networks: A Survey
- P-tensors: a General Framework for Higher Order Message Passing in Subgraph Neural Networks P-tensors: a General Framework for Higher Order Message Passing in Subgraph Neural Networks
- CIN++: Enhancing Topological Message Passing CIN++: Enhancing Topological Message Passing
- TMLR 2023: Self-Attention in Colors: Another Take on Encoding Graph Structure in Transformers Self-Attention in Colors: Another Take on Encoding Graph Structure in Transformers TMLR 2023
- ICML 2023: Equivariant Polynomials for Graph Neural Networks Equivariant Polynomials for Graph Neural Networks ICML 2023
- ICLR 2022: Boosting the Cycle Counting Power of Graph Neural Networks with I$^2$-GNNs Boosting the Cycle Counting Power of Graph Neural Networks with I$^2$-GNNs ICLR 2022
- Equivariant 3D-Conditional Diffusion Models for Molecular Linker Design Equivariant 3D-Conditional Diffusion Models for Molecular Linker Design
- TMLR 2022: How Expressive are Transformers in Spectral Domain for Graphs? How Expressive are Transformers in Spectral Domain for Graphs? TMLR 2022
- Meta-Aggregator: Learning to Aggregate for 1-bit Graph Neural Networks Meta-Aggregator: Learning to Aggregate for 1-bit Graph Neural Networks
- NeurIPS 2021: Weisfeiler and Lehman Go Cellular: CW Networks Weisfeiler and Lehman Go Cellular: CW Networks NeurIPS 2021
- NeurIPS 2019: Hyperbolic Graph Neural Networks Hyperbolic Graph Neural Networks NeurIPS 2019
- ICLR 2018: Syntax-Directed Variational Autoencoder for Structured Data Syntax-Directed Variational Autoencoder for Structured Data ICLR 2018
Graph regression on heavy-atom molecular graphs, predicting penalized logP (logP minus synthetic-accessibility and long-cycle penalties). Nodes are atom types and edges are bond types. Evaluated by MAE.
Dataset details
Task
Molecular property regression (constrained solubility). The primary catalog metric is MAE.
Data
Molecular property regression (constrained solubility).
Features
The representation follows the cited PyG loader or benchmark release. Node and edge fields are dataset-specific; preprocessing, graph direction, and any feature normalization must be kept fixed when reproducing a result.
Splits and evaluation
The catalog records these protocols or variants: Subset (12k), Full. Evaluation uses MAE (lower is better). Exact masks or folds must come from the cited release.
Quirks and pitfalls
The same short dataset name is used for incompatible processed artifacts and split protocols. Report the loader/version, directedness, feature preprocessing, and exact masks; transductive results should not be described as inductive.
Sources
250,000 train molecules. Much better quality expected compared to 12k subset.
SOTA Progression
new SOTA
matches SOTA at that time
other proposed result
other baseline result
Leaderboard
| # | Model | MAE ↓ | QED | RMSD | RMSE | Recovery | SA | Paper |
|---|---|---|---|---|---|---|---|---|
| 1 | TIGT | 0.0140
±
0.0010 | — | — | — | — | — | |
| 2 | FloydNet | 0.0160 | — | — | — | — | — | |
| 3 | GSSC | 0.0190
±
0.0010 | — | — | — | — | — | |
| 4 | PPGN++(6) | 0.0200
±
0.0010 | — | — | — | — | — | |
| 5 | PPGN++ | 0.0200
±
0.0010 | — | — | — | — | — | |
| 6 | CS-GNN | 0.0210
±
0.0010 | — | — | — | — | — | |
| 7 | CIN | 0.0220
±
0.0020 | — | — | — | — | — | |
| 8 | GNN-SSWL+ | 0.0220
±
0.0010 | — | — | — | — | — | |
| 9 | SSWL+ | 0.0220
±
0.0020 | — | — | — | — | — | |
| 10 | Subgraphormer | 0.0230
±
0.0010 | — | — | — | — | — | |
| 11 | GRIT | 0.0230
±
0.0010 | — | — | — | — | — | |
| 12 | GRIT 2-QiQRW | 0.0230
±
0.0020 | — | — | — | — | — | |
| 13 | I^2-GNN | 0.0230
±
0.0010 | — | — | — | — | — | |
| 14 | I^2-GNNs | 0.0230
±
0.0010 | — | — | — | — | — | |
| 15 | MPNNs | 0.0240
±
0.0010 | — | — | — | — | — | |
| 16 | First order P-tensors | 0.0240
±
0.0010 | — | — | — | — | — | |
| 17 | Local 2-GNN | 0.0240 | — | — | — | — | — | |
| 18 | ET+RRWP | 0.0240 | — | — | — | — | — | |
| 19 | SignNet | 0.0240
±
0.0030 | — | — | — | — | — | |
| 20 | Lorentz | 0.0240
±
0.0002 | — | — | — | — | — | |
| 21 | SUN | 0.0240
±
0.0030 | — | — | — | — | — | |
| 22 | DRFWL | 0.0250
±
0.0030 | — | — | — | — | — | |
| 23 | Graphormer | 0.0250
±
0.0040 | — | — | — | — | — | |
| 24 | GRIT 1-CQRW | 0.0250
±
0.0030 | — | — | — | — | — | |
| 25 | Graphormer-GD | 0.0250
±
0.0040 | — | — | — | — | — | |
| 26 | d-DRFWL(2) GNN | 0.0250
±
0.0030 | — | — | — | — | — | |
| 27 | ET | 0.0260
±
0.0030 | — | — | — | — | — | |
| 28 | Subgraph GNN | 0.0280 | — | — | — | — | — | |
| 29 | URPE | 0.0280
±
0.0020 | — | — | — | — | — | |
| 30 | Graphormer-URPE | 0.0280
±
0.0020 | — | — | — | — | — | |
| 31 | NGNN | 0.0290
±
0.0010 | — | — | — | — | — | |
| 32 | DSS-GNN | 0.0290
±
0.0030 | — | — | — | — | — | |
| 33 | ESAN | 0.0290
±
0.0030 | — | — | — | — | — | |
| 34 | Nested GIN | 0.0290
±
0.0010 | — | — | — | — | — | |
| 35 | CNNs | 0.0290
±
0.0010 | — | — | — | — | — | |
| 36 | MPNN | 0.0300 | — | — | — | — | — | |
| 37 | MAG-GNN | 0.0300
±
0.0020 | — | — | — | — | — | |
| 38 | Poincare | 0.0310
±
0.0001 | — | — | — | — | — | |
| 39 | GGNN | 0.0320
±
0.0020 | — | — | — | — | — | |
| 40 | Euclidean | 0.0330
±
0.0000 | — | — | — | — | — | |
| 41 | GNN-AK+ | 0.0340
±
0.0070 | — | — | — | — | — | |
| 42 | Graph ViT | 0.0350
±
0.0050 | — | — | — | — | — | |
| 43 | CRaWl | 0.0360
±
0.0050 | — | — | — | — | — | |
| 44 | SAT | 0.0360
±
0.0020 | — | — | — | — | — | |
| 45 | HIMP | 0.0360
±
0.0020 | — | — | — | — | — | |
| 46 | DTNN | 0.0400
±
0.0003 | — | — | — | — | — | |
| 47 | AgentNet | 0.0400
±
0.0060 | — | — | — | — | — | |
| 48 | MPNN | 0.0410
±
0.0002 | — | — | — | — | — | |
| 49 | delta-2-GNN | 0.0420
±
0.0030 | — | — | — | — | — | |
| 50 | Cy2C-GNNs | 0.0420
±
0.0010 | — | — | — | — | — |
SOTA performance
set a new SOTA at the time
matched the SOTA at the time