MOCK PDF — Annex IV §1–§9 placeholder
=====================================

NOTE: text-fixture mock. See `crates/silver-annex-iv-v2/MOCK.md`.

Model:        fixture-fine-tune-C
Variant:      narrow fine-tune over a synthetic-augmented corpus
Generated:    2026-05-04 (mock)
Spec ref:     EU AI Act Art. 11 + Annex IV; Art. 50 (transparency for
              synthetic content)

§1 — General description
------------------------
Fine-tune "fixture-fine-tune-C" extends fixture-foundation-model-A
on a corpus that mixes human-authored data and synthetic
augmentations from Model A itself (the parent foundation model).
Demonstrates the loop-detection T6+T8 refutation when synthetic
provenance is missing from the Croissant manifest.

§2 — Detailed description
-------------------------
§2(d) Datasheets: see `synthetic-data.json`. Synthetic-source flag
       missing on a subset of rows — exactly the case Art. 50 targets.

§3 — Monitoring, functioning and control
----------------------------------------
Continuous training-data ingest gate: refuse rows missing the
synthetic-source tag.

§4 — Performance metrics
------------------------
Validation perplexity: 7.9
Synthetic-tag completeness (post-gate): 100%

§5 — Risk-management system
---------------------------
Risk register flags model-collapse risk; quarterly synthetic-mix
ratio review.

§6 — Lifecycle changelog
------------------------
2026-04-20  Initial fine-tune.
2026-04-28  Re-training after synthetic-source gate enforced.

§7 — Harmonised standards applied
---------------------------------
ISO/IEC 23894:2023; AI Office Art. 53(1)(d) template.

§8 — Declaration of conformity
------------------------------
DoC reference: DOC-2026-MOCK-003.

§9 — Post-market monitoring plan
--------------------------------
Per-batch synthetic-source-tag completeness audit; quarterly
provenance-loop detection on ful