Open Graph Benchmark: Datasets for Machine Learning on Graphs
Presents OGB, a diverse suite of large-scale, realistic benchmark datasets with unified evaluation protocols for reproducible graph machine learning.
The Open Graph Benchmark (OGB) is a diverse set of challenging, realistic datasets for reproducible graph ML research. Datasets are large-scale (up to 100M+ nodes, 1B+ edges), span multiple tasks, and cover social, information, biological, molecular, source-code AST, and knowledge-graph domains. Each provides a unified evaluation protocol with application-specific splits and metrics. Experiments reveal scaling and out-of-distribution generalization challenges, and OGB offers an automated pipeline with public loaders and leaderboards.
Based on: Open Graph Benchmark: Datasets for Machine Learning on Graphs · Neural Information Processing Systems
Curated by Aramai Editorial
Read summary →