Matbench Discovery

Matbench Discovery is a benchmark that evaluates how accurately machine learning models predict the thermodynamic stability of inorganic crystalline materials. Modeled on high-throughput screening for new materials, it measures how much more efficiently a model identifies stable crystals than random selection would.

It is mainly used to compare machine learning interatomic potentials and property-prediction models in materials science.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.