Initial release validation — 2026-09-14
This historical report concerns the initial ieee-access-20260914-2897476
snapshot and its original reproduction-code version. The revised release
has separate checks in revised-reproduction.json and
revised-source-package.json; release.json identifies the current archive.
The standalone package was tested with a fresh CPython 3.11 environment and
the unchanged pinned publication requirements. pip check passed.
- 33 unit/regression tests passed.
- The frozen archive SHA-256 matched the release lock.
- All 19,226 evidence-manifest entries verified after extraction.
- Analysis reconstructed all 588 principal processes and the 16 separately counted historical source screens.
- All 40 regenerated CSV/JSON result files matched the archived references after normalizing only artifact-root path prefixes.
- All four publication figures were regenerated as PDF and PNG.
- During analysis, bubblewrap hid the original project and the directory holding precomputed reference profiles. No network-namespace isolation is claimed.
- 19,073 regular archive members were scanned for credential-like content. No model/optimizer checkpoint payloads were included.
The copied scientific source files are byte-identical to the original
289747694f1a71052acc7977e742bfc0a45286a2 snapshot; their individual hashes
are recorded in source-provenance.json. The new wrapper and download helper
are separate packaging code.
This is numerical/raw-evidence reproduction, not a new GPU experiment, learning-quality evaluation, author approval, license grant, or journal acceptance certification.