# Adversarial review log - rbp-aggregation-als-ftd (2026-10-09 revised edition)

**Authoring model:** Sol 6.1 - **Adversarial reviewer:** Claude Opus 5.5 (Anthropic) - **Date:** 2026-10-14 - Round 2

| Metric | Value |
|---|---|
| Level before / after | Γ / Γ (unchanged) |
| Confidence before / after | Moderate / Moderate (unchanged, re-derived after striking access depth) |
| Momentum before / after | Decisive work underway / **Mechanistic only** |
| Findings | 0 Critical - 2 Major - 1 Minor - 2 Editorial |
| Of which revision-introduced regressions | 1 (A1) |
| REFERENCES PMIDs verified | 63 / 63 (0 failed, 0 retracted) |
| Load-bearing claims spot-checked | 13 |
| Validator | PASS - 0 fail, 1 pre-existing WARN, 31 pass, validate_review.py 2.3 (this package's validator of record; see `validator.txt`) |

## Findings

| ID | Severity | Location | Original to corrected |
|---|---|---|---|
| A1 | Major (**revision-introduced regression**) | Spliced Consensus provenance block; `consensus-addendum.md` | "The earlier adversarial approval refers to its original edition" - false for this case; rescoped to record that no earlier adversarial approval is attached, because the original edition post-dates the 2026-10-06 pass. |
| A2 | Major (inherited) | AT A GLANCE table, M/S/D section, scorecard momentum | "Decisive work underway" justified by two completed publications, to "Mechanistic only" with the artifact rescoped to name the ongoing jacifusen trial and why it does not address the registered boundary. |
| A3 | Editorial (inherited) | review.md, 22 replacements, plus matching scorecard strings | "UpdateOf:"/"ErratumIn:"/"PMIDnot supplied" and missing spaces ("by29weeks", "20microgram") corrected. No numeral or PMID value changed. |
| A4 | Minor (inherited) | scorecard `final_score.confidence_rationale` | Band justified partly by reviewer access depth; restated on evidence density, coupled interventions and shared mouse lines. Band stands at Moderate. |
| A5 | Editorial | REFERENCES entry 24 (PMID 35075293) | Reference prints "Korben VA" (as PubMed renders it) while the prose says "Korobeynikov". Note added at the reference recording the discrepancy; neither form altered, because which is correct cannot be settled from the accessible record. |

Nine findings were checked and rejected, including the known Oct 9 reviewer-notes contradiction (which does **not** apply
to this case - its preserved notes never deny Consensus work), the scope-matched aggregation audit, the
`models_evaluated[0]` renderer-bug check and the 29-week survival figure verified against the original full text.
See `adversarial-findings.json`.

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