MatBench
Materials property prediction benchmark suite (used for some chemistry-adjacent ML).
Composite
83.3
Experimental validation
Retrospective
Stages
Developmental Candidate
Modalities
cross-modality
Task types
materials-property
Size
tasks: 13
samples: 132,000
samples: 132,000
License
MIT
First release
2020
Last updated
2025-02
Official site
Leaderboard
Dataset
→ dataset
Code / GitHub
HuggingFace
→ HF
Paper
Benchmarking materials property prediction methods: the Matbench test set and Automatminer reference algorithm · Dunn A, Wang Q, Ganose A, Dopp D, Jain A · 2020 · paper · doi:10.1038/s41524-020-00406-3 · 380 citations
Flags
none
Experts
Groups
Hosted by
Related benchmarks
—
Rubric (7-criterion)
rigor
5
coverage
3
maintenance
4
adoption
4
quality
5
accessibility
5
industry_relevance
3
Notes
Materials-science benchmark; relevant for formulation / co-crystal work.