PQC-Corpus: A Dataset of Linked Development Histories for Post-Quantum Cryptography

PQC-Corpus is a linked dataset of the development history of 179 public GitHub repositories about post-quantum cryptography (PQC), collected on 21–24 September 2026. It joins Git history with GitHub discussions: commits with their full patches, pull requests with their reviews, and issues with their comments, linked to each other where GitHub or Git records the link, so that standards updates, cryptographic review and migration work can be followed from discussion to code. Contents commits: 114,361 commits with full patches, changed files and the pull requests that contain them (47,729 in the filtered set) pull_requests: 11,919 pull requests with their commits, closed issues, comments, reviews and review comments issues: 5,713 issues with their comments and closing pull requests (5,687 in the filtered set) repositories: 179 repositories with their corpus and GitHub metadata Corpora A (81 repositories): at least 14.8% of their issues and pull requests name a PQC term; every commit, issue and pull request B (87): larger general-purpose projects; only issues, pull requests and commits that name one of 90 PQC terms, and commits of such pull requests C (11): repositories without issues or pull requests; every commit Full and filtered sets The full sets are everything collected. The filtered sets apply fixed, scripted rules: commits that reached the default branch or a merged pull request, each change once, without inherited fork history, non-PQC work in general-purpose libraries or housekeeping commits; issues without those opened by bots. Every removal is listed, with its reason, in the removal tables. Limitations Selection is by keyword: a record is not necessarily about PQC, and corpus B misses PQC work that names no term. Records are as of collection; later edits and deletions on GitHub are not reflected. Counts reflect the selection policy, not ecosystem-wide PQC adoption. Personal data and licence Usernames and Git author names are kept as published on GitHub. Author e-mail addresses are not included; commits carry only a keyed hash of them (addresses written inside messages and patches are unchanged). The selection, structure and documentation are licensed under CC BY 4.0; the content of the records remains under the source repositories' licences. Links Code that collects and builds the dataset: github.com/SouhardyaSaha/pqc_corpus Same release on Hugging Face: pqc-datasets/pqc-corpus