Scaling Reproducibility: An AI-Assisted Workflow for Large-Scale Reanalysis

• https://doi.org/10.31235/osf.io/ru5fa_v1
• Yiqing Xu, Leo Yang

Reproducibility, AI for science, Instrumental variable

AI-assisted workflow for large-scale reanalysis. Combines fixed diagnostic templates with version-controlled code to standardize replication materials across heterogeneous studies. Tested on 92 instrumental variable studies: 87% end-to-end success overall, 100% reproducibility conditional on accessible data and code.

Related: OSC2015estimating, Bak Coleman2022replication, Protzko2023high, Hoyle2021automated