Graph machine learning results organized by dataset in live leaderboard
A Living Leaderboard for Graph Machine Learning
Machine Learning
Summary
It is hard to tell if new graph machine learning methods are really better because results are scattered across many papers using different data and comparisons. The authors created an online leaderboard that collects and organizes thousands of results by dataset, so people can see how new methods compare to past ones clearly. They found many papers still use old datasets and report baseline results that are much worse than known results. This leaderboard helps improve how graph machine learning progress is measured.
What this means in practice
- •For machine learning engineers: Quickly compare new graph learning models against a large, organized set of previous results by dataset to validate improvements.
- •For data science teams: Standardize evaluation practices by referencing a centralized leaderboard that highlights outdated baselines and consolidates benchmarking results.
Authors
Martin Ritzert
Abstract
Graph machine learning benchmarking results are scattered across papers that use quite different sets of datasets and baseline methods, making it difficult to tell whether a reported gain reflects architectural progress or merely an outdated comparison or weakly tuned baseline. We introduce a living, graph-learning-specific leaderboard that organizes reported results by dataset, putting new results into context. As of 15 October 2026 our LLM-based extraction pipeline extracted 27,723 eligible results across 111 datasets from 885 peer-reviewed papers. Our analysis of these results quantifies concerns about graph-learning benchmarking: legacy datasets are still actively used and many result tables contain baseline results that are drastically lower than those already available in the literature, e.g. 45% Amazon-ratings tables and 37% for Peptides-struct. The leaderboard allows to quickly compare new results against compatible prior results, aiming to improve the overall quality of graph learning benchmarking.