GitPedia

Lrgb

Long Range Graph Benchmark, NeurIPS 2022 Track on D&B

From vijaydwivedi75·Updated May 13, 2026·View on GitHub·

We present the **Long Range Graph Benchmark (LRGB)** with 5 graph learning datasets that arguably require long-range reasoning to achieve strong performance in a given task. - PascalVOC-SP - COCO-SP - PCQM-Contact - Peptides-func - Peptides-struct The project is written primarily in Jupyter Notebook, distributed under the MIT License license, first published in 2022. Key topics include: graph-benchmark, graph-datasets, graph-neural-networks, graph-representation-learning, graph-transformer.

LRGB: Long Range Graph Benchmark

arXiv

<img src="https://i.imgur.com/2LKoGbu.png" align="right" width="275"/>

We present the Long Range Graph Benchmark (LRGB) with 5 graph learning datasets that arguably require
long-range reasoning to achieve strong performance in a given task.

  • PascalVOC-SP
  • COCO-SP
  • PCQM-Contact
  • Peptides-func
  • Peptides-struct

In this repo, we provide the source code to load the proposed datasets and run baseline experiments.
The repo is based on GraphGPS which is built using PyG and GraphGym from PyG2.

Update: Reassessment of LRGB

For a reassessment of the baselines on which LRGB were initially evaluated, we refer to this paper and thank @toenshoff for the PR on PCQM-Contact's evaluation metric.

Overview of Datasets

DatasetDomainTaskNode Feat. (dim)Edge Feat. (dim)Perf. Metric
PascalVOC-SPComputer VisionNode PredictionPixel + Coord (14)Edge Weight (1 or 2)macro F1
COCO-SPComputer VisionNode PredictionPixel + Coord (14)Edge Weight (1 or 2)macro F1
PCQM-ContactQuantum ChemistryLink PredictionAtom Encoder (9)Bond Encoder (3)Hits@K, MRR
Peptides-funcChemistryGraph ClassificationAtom Encoder (9)Bond Encoder (3)AP
Peptides-structChemistryGraph RegressionAtom Encoder (9)Bond Encoder (3)MAE

Statistics of Datasets

Dataset# Graphs# Nodesμ Nodesμ Deg.# Edgesμ Edgesμ Short. Pathμ Diameter
PascalVOC-SP11,3555,443,545479.405.6530,777,4442,710.4810.74±0.5127.62±2.13
COCO-SP123,28658,793,216476.885.65332,091,9022,693.6710.66±0.5527.39±2.14
PCQM-Contact529,43415,955,68730.142.0332,341,64461.094.63±0.639.86±1.79
Peptides-func15,5352,344,859150.942.044,773,974307.3020.89±9.7956.99±28.72
Peptides-struct15,5352,344,859150.942.044,773,974307.3020.89±9.7956.99±28.72

Python environment setup with Conda

bash
conda create -n lrgb python=3.9
conda activate lrgb

conda install pytorch=1.9 torchvision torchaudio -c pytorch -c nvidia
conda install pyg=2.0.2 -c pyg -c conda-forge
conda install pandas scikit-learn

# RDKit is required for OGB-LSC PCQM4Mv2 and datasets derived from it.  
conda install openbabel fsspec rdkit -c conda-forge

# Check https://www.dgl.ai/pages/start.html to install DGL based on your CUDA requirements
pip install dgl-cu111 dglgo -f https://data.dgl.ai/wheels/repo.html

pip install performer-pytorch
pip install torchmetrics==0.7.2
pip install ogb
pip install wandb

conda clean --all

Running GraphGPS

bash
conda activate lrgb

# Running GCN baseline for Peptides-func.
python main.py --cfg configs/GCN/peptides-func-GCN.yaml  wandb.use False

# Running SAN baseline for PascalVOC-SP.
python main.py --cfg configs/SAN/vocsuperpixels-SAN.yaml  wandb.use False

The scripts for all experiments are located in run directory.

W&B logging

To use W&B logging, set wandb.use True and have a gtransformers entity set-up in your W&B account (or change it to whatever else you like by setting wandb.entity).

<!-- ### Datasets Links Following is a list of direct URL links of the proposed datasets that are used in the respective files in the [graphgps/loader/dataset](graphgps/loader/dataset) directory. Note that the links below are just for information and there is no need to manually download these source files. The files are automatically downloaded in the corresponding files in the [graphgps/loader/dataset](graphgps/loader/dataset) directory. 1. PascalVOC-SP: [link](https://www.dropbox.com/s/8x722ai272wqwl4/voc_superpixels_edge_wt_region_boundary.zip?dl=0) 2. COCO-SP: [link](https://www.dropbox.com/s/r6ihg1f4pmyjjy0/coco_superpixels_edge_wt_region_boundary.zip?dl=0) 3. PCQM-Contact: [link](https://datasets-public-research.s3.us-east-2.amazonaws.com/PCQM4M/pcqm4m-contact.tsv.gz) 4. Peptides-func: [link](https://www.dropbox.com/s/ol2v01usvaxbsr8/peptide_multi_class_dataset.csv.gz?dl=0) 5. Peptides-struct: [link](https://www.dropbox.com/s/464u3303eu2u4zp/peptide_structure_dataset.csv.gz?dl=0) -->

License Information

DatasetDerived fromOriginal LicenseLRGB Release License
PascalVOC-SPPascal VOC 2011Custom*Custom*
COCO-SPMS COCOCC BY 4.0CC BY 4.0
PCQM-ContactPCQM4Mv2CC BY 4.0CC BY 4.0
Peptides-funcSATPdbCC BY-NC 4.0CC BY-NC 4.0
Peptides-structSATPdbCC BY-NC 4.0CC BY-NC 4.0

*Custom License for Pascal VOC 2011 (respecting Flickr terms of use)

Leaderboards

The leaderboards of various models' performance on the datasets in LRGB are at paperswithcode.

Currently reported results (last update on Aug 10th, 2023)

<details> <summary> PascalVOC-SP (Node Classification) </summary>
ModelTest F1 (higher is better)Reference#params
Exphormer0.3975±0.0037Shirzad, Velingker, Venkatachalam, et al, ICML 2023509k
GraphGPS0.3748±0.0109Rampášek et al, NeurIPS 2022510k
Cache-GNN+LapPE0.3462±0.0085Ma et al, KDD 2023500k
DRew-GatedGCN+LapPE0.3314±0.0024Gutteridge et al, ICML 2023502k
SAN+LapPE0.3230±0.0039Dwivedi et al, NeurIPS 2022531k
SAN+RWSE0.3216±0.0027Dwivedi et al, NeurIPS 2022468k
GatedGCN+LapPE+virtual node0.3103±0.0068Cai et al, ICML 2023502k
GatedGCN0.2873±0.0219Dwivedi et al, NeurIPS 2022502k
GatedGCN+LapPE0.2860±0.0085Dwivedi et al, NeurIPS 2022502k
Transformer+LapPE0.2694±0.0098Dwivedi et al, NeurIPS 2022501k
GCNII0.1698±0.0080Dwivedi et al, NeurIPS 2022492k
GCN0.1268±0.0060Dwivedi et al, NeurIPS 2022496k
GINE0.1265±0.0076Dwivedi et al, NeurIPS 2022505k
</details> <details> <summary> COCO-SP (Node Classification) </summary>
ModelTest F1 (higher is better)Reference#params
Exphormer0.3455±0.0009Shirzad, Velingker, Venkatachalam, et al, ICML 2023499k
GraphGPS0.3412±0.0044Rampášek et al, NeurIPS 2022516k
Cache-GNN+LapPE0.2793±0.0033Ma et al, KDD 2023500k
GatedGCN0.2641±0.0045Dwivedi et al, NeurIPS 2022509k
Transformer+LapPE0.2618±0.0031Dwivedi et al, NeurIPS 2022508k
SAN+LapPE0.2592±0.0158Dwivedi et al, NeurIPS 2022536k
GatedGCN+LapPE0.2574±0.0034Dwivedi et al, NeurIPS 2022509k
SAN+RWSE0.2434±0.0156Dwivedi et al, NeurIPS 2022474k
GCNII0.1404±0.0011Dwivedi et al, NeurIPS 2022505k
GINE0.1339±0.0044Dwivedi et al, NeurIPS 2022515k
GCN0.0841±0.0010Dwivedi et al, NeurIPS 2022509k
</details> <details> <summary> Peptides-func (Graph Classification) </summary>
ModelTest AP (higher is better)Reference#params
DRew-GCN+LapPE0.7150±0.0044Gutteridge et al, ICML 2023502k
GRIT0.6988±0.0082Ma, Lin, et al, ICML 2023443k
GraphMLP-Mixer0.6970±0.0080He et al, ICML 2023397k
Graph ViT0.6942±0.0075He et al, ICML 2023692k
MGT+WavePE0.6817±0.0064Ngo, Hy, et al, 2023499k
PathNN0.6816±0.0026Michel, Nikolentzos et al, ICML 2023510k
GatedGCN+RWSE+virtual node0.6685±0.0062Cai et al, ICML 2023506k
Cache-GNN+LapPE0.6671±0.0056Ma et al, KDD 2023500k
Graph Diffuser0.6651±0.0010Glickman & Yahav, 2023509k
CIN++0.6569±0.0117Giusti et al, 2023~500k
GraphGPS0.6535±0.0041Rampášek et al, NeurIPS 2022504k
Exphormer0.6527±0.0043Shirzad, Velingker, Venkatachalam, et al, ICML 2023446k
SAN+RWSE0.6439±0.0075Dwivedi et al, NeurIPS 2022500k
SAN+LapPE0.6384±0.0121Dwivedi et al, NeurIPS 2022493k
Transformer+LapPE0.6326±0.0126Dwivedi et al, NeurIPS 2022488k
GatedGCN+RWSE0.6069±0.0035Dwivedi et al, NeurIPS 2022506k
GCN0.5930±0.0023Dwivedi et al, NeurIPS 2022508k
GatedGCN0.5864±0.0077Dwivedi et al, NeurIPS 2022509k
GCNII0.5543±0.0078Dwivedi et al, NeurIPS 2022505k
GINE0.5498±0.0079Dwivedi et al, NeurIPS 2022476k
</details> <details> <summary> Peptides-struct (Graph Regression) </summary>
ModelTest MAE (lower is better)Reference#params
Cache-GNN+LapPE0.2358±0.0013Ma et al, KDD 2023500k
Graph ViT0.2449±0.0016He et al, ICML 2023561k
MGT+WavePE0.2453±0.0025Ngo, Hy, et al, 2023499k
GRIT0.2460±0.0012Ma, Lin, et al, ICML 2023439k
Graph Diffuser0.2461±0.0010Glickman & Yahav, 2023509k
Exphormer0.2481±0.0007Shirzad, Velingker, Venkatachalam, et al, ICML 2023426k
GCN+virtual node0.2488±0.0021Cai et al, ICML 2023508k
Graph MLP-Mixer0.2494±0.0007He et al, ICML 2023397k
GraphGPS0.2500±0.0005Rampášek et al, NeurIPS 2022504k
CIN++0.2523±0.0013Giusti et al, 2023~500k
Transformer+LapPE0.2529±0.0016Dwivedi et al, NeurIPS 2022488k
DRew-GCN+LapPE0.2536±0.0015Gutteridge et al, ICML 2023495k
SAN+RWSE0.2545±0.0012Dwivedi et al, NeurIPS 2022500k
PathNN0.2545±0.0032Michel, Nikolentzos et al, ICML 2023469k
NPQ+GATv20.2589±0.0031Jain et al, KLR Workshop at ICML, 2023NA
SAN+LapPE0.2683±0.0043Dwivedi et al, NeurIPS 2022493k
GatedGCN+RWSE0.3357±0.0006Dwivedi et al, NeurIPS 2022506k
GatedGCN0.3420±0.0013Dwivedi et al, NeurIPS 2022509k
GCNII0.3471±0.0010Dwivedi et al, NeurIPS 2022505k
GCN0.3496±0.0013Dwivedi et al, NeurIPS 2022508k
GINE0.3547±0.0045Dwivedi et al, NeurIPS 2022476k
</details> <details> <summary> PCQM-Contact (Link Prediction) </summary>
ModelTest MRR (higher is better)Test Hits@1Test Hits@3Test Hits@10Reference#params
Exphormer0.3637±0.0020Shirzad, Velingker, Venkatachalam, et al, ICML 2023396k
Cache-GNN+RWSE0.3488±0.00080.1463±0.00110.4102±0.00080.8693±0.0008Ma et al, KDD 2023500k
DRew-GCN0.3444±0.0017Gutteridge et al, ICML 2023515k
Graph Diffuser0.3388±0.00110.1369±0.00120.4053±0.00110.8592±0.0007Glickman & Yahav, 2023521k
SAN+LapPE0.3350±0.00030.1355±0.00170.4004±0.00210.8478±0.0044Dwivedi et al, NeurIPS 2022499k
SAN+RWSE0.3341±0.00060.1312±0.00160.4030±0.00080.8550±0.0024Dwivedi et al, NeurIPS 2022509k
GraphGPS0.3337±0.0006Rampášek et al, NeurIPS 2022513k
GatedGCN+RWSE0.3242±0.00080.1288±0.00130.3808±0.00060.8517±0.0005Dwivedi et al, NeurIPS 2022524k
GCN0.3234±0.00060.1321±0.00070.3791±0.00040.8256±0.0006Dwivedi et al, NeurIPS 2022504k
GatedGCN0.3218±0.00110.1279±0.00180.3783±0.00040.8433±0.0011Dwivedi et al, NeurIPS 2022527k
GINE0.3180±0.00270.1337±0.00130.3642±0.00430.8147±0.0062Dwivedi et al, NeurIPS 2022517k
Transformer+LapPE0.3174±0.00200.1221±0.00110.3679±0.00330.8517±0.0039Dwivedi et al, NeurIPS 2022502k
GCNII0.3161±0.00040.1325±0.00090.3607±0.00030.8116±0.0009Dwivedi et al, NeurIPS 2022501k
</details>

Citation

If you find this work useful, please cite our paper:

bibtex
@inproceedings{dwivedi2022LRGB,
  title={Long Range Graph Benchmark}, 
  author={Dwivedi, Vijay Prakash and Rampášek, Ladislav and Galkin, Mikhail and Parviz, Ali and Wolf, Guy and Luu, Anh Tuan and Beaini, Dominique},
  booktitle={Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
  year={2022},
  url={https://openreview.net/forum?id=in7XC5RcjEn}
}

Contributors

Showing top 4 contributors by commit count.

vijaydwivedi75

vijaydwivedi75

6 commits

migalkin

migalkin

2 commits

toenshoff

toenshoff

2 commits

rampasek

rampasek

1 commits

View all contributors on GitHub →

This article is auto-generated from vijaydwivedi75/lrgb via the GitHub API.Last fetched: 6/20/2026