About MLCommons® Science

The MLCommons Science Working Group is dedicated to advancing the frontier of scientific machine learning. Our primary goal is to establish standardized benchmarks that enable the community to evaluate, compare, and improve AI models applied to complex scientific problems.

By creating a transparent and reproducible framework for benchmarking, we aim to accelerate scientific discovery and ensure that AI tools are developed with the rigor required for scientific research.

Official Resources

Stay connected with our official governance and guidelines:

• Official Working Group Page: mlcommons.org/en/groups/research-science/
• Science Policy Document: View the official policy on GitHub
• Call for Benchmarks: Learn how to submit a new scientific benchmark

Benchmark Catalog & Tools

We provide centralized resources to help researchers discover and utilize scientific benchmarks:

• Benchmark Catalog: A filterable and searchable index of scientific benchmarks. Explore the Catalog
• SABATH: The Scientific AI Benchmark and Tool Hub for accessing shared benchmarking tools. Visit SABATH on GitHub

Current Benchmarks

The following benchmarks are currently available, including descriptions and source code:

• Cloudmask: View on GitHub
• Earthquake: View on GitHub
• Stemdl: View on GitHub
• Uno: View on GitHub

Community & Contribution

We welcome contributions from the global scientific and AI communities.

• GitHub Repository: mlcommons/science
• Contributor Guidelines: See our Contributing Page to learn how to get involved.
• News & Tags: Explore our tags page for categorized updates.