Research · the paper
Is there a genuine gap in CI/CD regression testing for quantum software?
A primary-source-verified prior-art study · Said Mohaddes Sadeqi · v1.1.1, August 2026
Abstract
This study asks whether an open-source, pytest-native infrastructure that runs quantum software projects across SDK versions and SDKs and automatically detects regressions or circuit-equivalence failures already exists. Starting from a hypothesis that none did, 14 claims were checked against primary sources, official repositories, raw CI configurations, arXiv full texts and DOI records, that were opened and inspected rather than accepted from search snippets; 22 prior-art systems were catalogued. The evidence does not support the original hypothesis: cross-version regression testing with CI output exists (QUTest, 2026, Qiskit-only); a pytest-native harness across eight SDKs exists for benchmarking (Benchpress) but detects nothing; pytest equivalence plugins exist but are immature; and a bug corpus (Bugs4Q) exists, contrary to the starting assumption. The narrowest defensible gap is a pytest-native, cross-SDK, automated equivalence/regression-detection framework packaged as reusable CI tooling, addressed by the companion tool released with this record. Two starting assumptions found false are reported as such; limitations are stated.
How to cite
@misc{sadeqi2026quantumcicd,
author = {Sadeqi, Said Mohaddes},
title = {Is there a genuine gap in CI/CD regression testing for quantum software?
A primary-source-verified prior-art study},
year = {2026},
month = aug,
version = {1.1.1},
howpublished = {Research record and paper, GitHub release v1.1.1},
url = {https://github.com/sadeqisaidmohaddes-star/quantum-cicd-research/releases/tag/v1.1.1},
note = {Companion tool: pytest-qequiv. DOI via Zenodo archival pending.}
}A CITATION.cff in the repository lets GitHub generate APA/BibTeX directly.
Archival status
- ✓ Versioned GitHub release with the PDF attached and
CITATION.cff+.zenodo.jsonin the tree. - ○ Zenodo DOI: the repository is prepared for Zenodo's GitHub archival; the DOI is minted when the owner enables the integration and re-publishes the release. Until then this page shows no DOI rather than a placeholder.
- ○ arXiv: not submitted at the check date. The paper states its limitations (English-language search, no empirical prototype comparison) that a reviewer would ask about first.
What the paper is generated from
The PDF is rendered from the research record: the claims table, the prior-art matrix, the gap analysis and the sources list are the source of truth, and continuous integration fails if the rendered claims drift from evidence.json. Every reference in the paper was opened during the study.