Should Exist cataloguePublic repositories for transparent model evaluationAI and statistical models used for public decisions should have reproducible evaluation sets, limitations, and independent test results.