SQL Database for the Replication Package of "When AI Models Become Dependencies: Studying the Evolution of Pre-Trained Model Reuse in Downstream Software Systems"

This Zenodo record provides the relational SQL database for the replication package of the study, "When AI Models Become Dependencies: Studying the Evolution of Pre-Trained Model Reuse in Downstream Software Systems", submitted to IEEE Transactions on Software Engineering (TSE). The dataset captures downstream repositories that reuse pre-trained models (PTMs), together with release history, branch/release-line structure, static-analysis evidence, PTM snapshots, and PTM change outcomes across releases. The schema includes repository metadata (downstream_repos, downstream_repo_info, owners), software evolution traces (commits, issues, pull_requests, releases, tags, discussions), and analysis-specific tables for PTM reuse extraction (reused_files, signatures, reused_files_signature_matches, reused_files_to_models, reused_files_model_snapshots_lists_for_releases) and release-level change information (release_lines, release_line_model_changes). This release is intended for transparent, reproducible re-analysis without requiring full API recollection from GitHub/Hugging Face, which may drift over time. The dump is designed to be used with the code and scripts in the companion GitHub replication package.