Release
Evaluation reports, strict proof packs, and Transformers v5
A terminology reset (report/evaluate), stricter proof-pack verification, and a clean upgrade path for Hugging Face Transformers v5.
Release: InvarLock 0.3.8 - Report/evaluate terminology and stricter pack verification
Highlights
- “Certificate” is now “report” everywhere (and
evaluatereplacescertify). - Proof packs can now verify in strict mode (fail-closed when signatures/contents don’t match expectations).
- Runtime summaries, version flags, Transformers v5 loading, and report bundle filenames are aligned.
This release is mostly about making the story of an InvarLock run easier to understand for both operators and reviewers. The breaking cleanup is broad: “certificate” becomes report across artifacts, docs, scripts, notebooks, and Python API surfaces, while CLI terminology converges on evaluate instead of certify.
Proof packs get stricter evidence handling. verify_pack.sh --strict and PACK_STRICT_MODE=1 can fail closed on missing or invalid GPG signatures, unexpected pack contents, checksum tampering, and extra files. Manifests now bind the checksum digest and can record the signing key fingerprint, which makes “what exactly did I verify?” a more answerable question.
On the ecosystem side, this release adds --version / -V, includes runtime and confidence-interval information in evaluate summaries, updates bundle filenames to evaluation.report.json and evaluation_report.md, rejects legacy Transformers v4 load keys, and migrates loading contracts for Transformers v5. It also reduces noisy Hugging Face output in CI/release profiles and keeps calibrate import-safe for docs/example validation without torch installed.
For the immutable release record, read the tagged CHANGELOG.md for v0.3.8.
More from the blog
Continue through recent releases and implementation notes.
Synthesis
The Minimum Evidence Surface for Trustworthy Weight-Edit Results
A trustworthy weight-edit result needs more than a benchmark delta. It needs a bounded claim, an exactly paired comparison, and verification that rejects incomplete evidence.
Release
Evidence packs and explicit runtime provenance
InvarLock 0.8.0 moves the public bundle surface to evidence packs, pins docs to versioned release paths, and makes container-vs-host runtime provenance explicit across evaluate and verify.
Research Note
Fail-Closed Verification for Weight-Edit Evaluation
A verifier is only useful if it rejects incomplete evidence. InvarLock's verification path is designed to stop stronger claims when the evidence bundle is missing or inconsistent.