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

Summary

Social-science (not ML) demonstration that an AI-assisted pipeline can retrieve materials, rebuild environments, execute code and match outputs to published regression-table point estimates at scale: 384 stratified studies from the top three political science journals (2010-2025), 3,523 empirical models replicated. Reproducibility rises from 20.8% pre-DA-RT to 82.5% after, and 92.1% conditional on an accessible replication package; the same infrastructure then supports systematic IV re-diagnostics across 597 specifications. The closest existing template for an IS/econ replication-agent argument.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a high-priority candidate for the RISE knowledge base (angle: replication and reproducibility). Verified against: arXiv abstract page opened 2026-08-29 (title, 2 authors, submitted 17 Feb 2026, full abstract read)

Critique / open questions

Queued — not yet assessed.

Key quotes

Computational reproducibility is central to scientific credibility, yet verifying published results at scale remains costly. (abstract)

We develop an AI-assisted workflow for automated full-paper replication -- retrieving materials, reconstructing environments, executing code, and matching outputs to point estimates reported in regression tables. (abstract)