Replication package for "Small Language Models on Physical Edge Devices: A Systematic Review of Capability and Device-Resource Evidence"

This package accompanies a systematic review (PRISMA 2020) of studies that run small language models on physical edge hardware, examining capability and device-resource evidence and how completely it is reported. It contains the extraction workbook (original and corrected), the register of 393 source-audit findings with a cell-level log of every correction, a random-sample re-extraction, a blinded human validation, the search manual, the Python scripts that regenerate every count and figure in the paper (run ./run_all.sh), and the eight figures. Extraction was AI-assisted (Claude Opus 5.5 via Manus); error rates are reported in the README. The source papers are not redistributed; each is identified by DOI in the workbook. Scripts are released under the MIT licence; all other content under CC BY 4.0.