# Adversarial review log — carbohydrate-insulin-obesity

**Authoring model:** Sol 6.1 (recorded in the score card as OpenAI Codex with HSEF v5.4; Claude red-team synthesis
2026-09-30). **Adversarial reviewer:** Claude Opus 5.5 (Anthropic). **Date:** 2026-10-14.
**Round:** 2 (2026-10-14). **Canonical path:** `consensus_consistency_audit_2026-10-09/revised/carbohydrate-insulin-obesity/carbohydrate-insulin-obesity.review.md`.

## Summary

| Level before | Level after | Confidence before | Confidence after | Critical | Major | Minor | Editorial | Approval |
|---|---|---|---|---|---|---|---|---|
| Γ — Mechanistically plausible, partial system support | Γ (unchanged) | High | High (checked against lead policy; no access-depth reasoning present, nothing to re-derive) | 0 | 1 | 2 | 0 | Approved as revised |

PMIDs checked 44/44, all resolved to the stated first author, title, journal, year and DOI. Seven linked correction
notices confirmed to exist. Retraction / expression-of-concern screen: none found. Load-bearing claims spot-checked:
12. Validator (archived HSEF v5.4 `validate_review.py`, reports 2.4): **pass** — 0 FAIL, **0 WARN** (down from 5),
33 PASS.

This was the strongest-built review in the group. Reference hygiene was essentially perfect, every rubric and cap
quotation was verbatim and correctly attributed, and the three defects the round-2 brief lists as recurring
(glued numerals, empty DOI fields, "PMIDnot supplied") were all absent. The one substantive finding is a spliced
provenance sentence.

## Findings

| ID | Severity | Location | Original → corrected | Evidence |
|---|---|---|---|---|
| CIO-1 | Major | Consensus provenance reconciliation block | "The earlier adversarial approval refers to its original edition" → a sentence naming what the 2026-10-06 approval actually covered | Misattribution rather than fabrication. An adversarial approval dated 2026-10-06 does exist for this topic, but it covered the separate prepublication **research-update addendum** (8 timeline identifiers, 4 gap statements, 4 correction notices — `prepublication-enrichment-2026-10-05/website-series/carbohydrate-insulin-obesity/review-addendum.md` and the matching website record). This review document, with 44 references and 31 registered claims, was never in that approval's scope. As written the sentence told a reader this review had already been adversarially approved. |
| CIO-2 | Minor | `citation_ledger` rows 26, 28, 29, 30, 31 | `access_depth: "metadata_only"` → `abstract_metadata` | `metadata_only` is not one of the three framework-defined tokens, which is what raised the five `v5.access_depth` validator warnings. The rows are the correspondence records, which § 9 says "carry only their existence and stated position" — abstract/metadata depth. Vocabulary fix only; it claims no reading that was not done. All five warnings cleared. |
| CIO-3 | Minor | VALIDATION NOTE | Validator versions transposed and a non-existent file named → corrected to the actual chronology and to `validation-report.json` | Validator 2.2 was used in the 2026-09-30 red-team pass (`redteam/REDTEAM_REPORT.md`); the archived HSEF v5.4 validator, which reports 2.4, was used later. No file named `carbohydrate-insulin-obesity.validation.txt` exists in this package or in the review's working tree. |

## Rejected findings (checked, no change)

1. **Confidence High beside a Γ letter.** Checked against the framework and the lead's two-step test and upheld. The
   rationale makes no appeal to reviewer access depth; it rests on two opposing syntheses of 29 and 32
   controlled-feeding studies, four large independent long-term RCTs, a 53-trial meta-analysis, inpatient trials from
   more than one laboratory and an independent free-living replication. The framework explicitly contemplates
   (Γ, high). Nothing to re-derive.
2. **One print-vs-e-pub year difference of one** (Soto-Mota, J Nutr 2024 vs 2023). Standing convention.
3. **"Best-supported rival" renderer defect.** Rank 1 (integrated energy homeostasis, M2) is `is_nominal: false`;
   the nominal M1 ranks 4. Defect absent.
4. **Collective author.** Reference 33 renders the DIOGENES consortium correctly ("…et al.; Diogenes Project"), with
   no blank author and no stray leading comma. The only collective author in the group, handled correctly.
5. **Rendering defects.** No glued numerals, no empty `DOI:` field, no "PMIDnot supplied" string.
6. **REVIEWER NOTES vs spliced Consensus section.** No "no Consensus work was performed" statement exists here;
   nothing contradicts the Consensus block.
7. **Momentum.** Already "Mechanistic only", and it correctly treats a trial that has published only its design as
   not satisfying "Decisive work underway".
8. **Dated-fact check.** A fresh PubMed search on 2026-10-14 confirms the residential isocaloric trial's primary
   results are still unpublished, so the review's claim holds at its cutoff and today.

## Residual limitations

Unchanged from the review's own statement: refs 3 and 7 could not be read in full; refs 8, 15, 17 and 22–44 were read
at abstract depth. No rubric in the packet covers body-weight or energy regulation; rubric 06's thresholds remain a
declared adaptation. This pass re-derived no number from raw trial data.

## Changed files

| File | sha256 | bytes |
|---|---|---|
| carbohydrate-insulin-obesity.review.md | b5d08268df0dbc02ba76f4448e7b6b75bfab0dbd642ea60f65b83352979bb548 | 78381 |
| carbohydrate-insulin-obesity.scorecard.json | 390442b5c6aaca273a3fe6f3bf67c5a6a88341b57ed9432e720d23640fedbe64 | 165002 |

`SHA256.txt` regenerated; `validation-report.json` re-run (0 FAIL, 0 WARN, 33 PASS). Pre-edit copies in
`sol61-original/`.

## For the lead

No level, confidence, role or M/S/D value changed, so no package-level scorecard row needs editing. One note worth
propagating: the 2026-10-06 adversarial approval recorded for this topic covers the prepublication addendum, not the
review, and any index that lists this slug as "adversarially approved 2026-10-06" should be read accordingly.


## Collective-author screen (both known forms)

Screened for a dropped author leaving a stray leading comma, for a consortium name truncated to its last word so
that it reads as an ordinary surname ("Group et al.", "Network et al.", "Consortium et al."), and for dropped
nobiliary particles. Method: every reference's rendered first author compared against the PubMed record,
including records whose first author object comes back empty because the article is credited to a consortium.
Result: one collective author — reference 33, the DIOGENES consortium, rendered as "Larsen TM, Dalskov SM, van Baak M, et al.; Diogenes Project", with the consortium present and appended rather than dropped or truncated. Particles intact ("Soto-Mota", "van Baak" mid-list). No defect.

## Line-ending note

The edit pass initially normalised these files from CRLF to LF. That was unintended and has been reversed: the
review and score card are back on CRLF, matching their pre-edit state and the rest of the package, and the hashes
below are the final CRLF files.

**Approval:** Following adversarial review and the corrections listed above, the revised document is approved. — Claude Opus 5.5 (Anthropic), 2026-10-14.