Reproducibility Corpus

This dataset accompanies the Semantic Reproducibility Engine (SRE) project, developed as part of the NII Open Collaborative Research programme. It provides a curated, stratified corpus of 95 machine learning repositories drawn from Papers with Code, each characterised by a set of reproducibility indicators measuring the presence or absence of environment specifications, dependency declarations, random seed documentation, and containerisation support. The dataset is designed to serve as a benchmark for evaluating automated reproducibility auditing tools, FAIR assessment frameworks, and metadata completeness scoring models in computational science.