# Adversarial review log — 3d-genome-regulation (revised edition)

Authoring model: **Sol 6.1**. Adversarial reviewer: **Claude Opus 5.5 (Anthropic)**. Date: **2026-10-14**.
Round 2; group 7 of `NEW_REVIEW_INVENTORY_2026-10-14.json`. Canonical copy is the 9 October 2026 revised edition.

## Summary

| Item | Value |
|---|---|
| Level before | Γ — Mechanistically plausible, partial system support / Moderate |
| Level after | Γ — Mechanistically plausible, partial system support / Moderate (unchanged) |
| Findings: Critical | 0 |
| Findings: Major | 2 |
| Findings: Minor | 3 |
| Findings: Editorial | 1 |
| Rejected findings | 7 |
| PMIDs checked / failed | 60 / 0 |
| Load-bearing claims spot-checked | 11 |
| Validator | FAIL (1 fail, 1 warn, 33 pass) — `v5.lay_summary` FAIL and `v5.claim_form` WARN; identical before and after. |
| Approval | **Approved as revised.** |

## Findings

| ID | Severity | Location | Original → corrected | Basis |
|---|---|---|---|---|
| AR-1 | Major | AT A GLANCE table; final-score block; score card momentum | Momentum "Decisive work underway" → "Mechanistic only"; artifact rewritten | The artifact itself says the named work "broadens scope without a topology-only primacy test" and claims no live trial, and the Stability entry says no imminent result crossing broad topology primacy was identified. By the framework's table that is "Mechanistic only". Momentum modifies no letter. |
| AR-2 | Minor | Final-score block Divergence line; score card divergence justification | Artifact added naming the two conflicting framings (field framing of folding as a primary determinant versus the acute-depletion literature's own "expression largely maintained" framing) | Divergence requires a checkable artifact; both prose and card asserted only its absence. Value unchanged at Contested. |
| AR-3 | Minor | Evidence card [42]; two score-card ledger rows | "ErratumIn: PMIDnot supplied" → "ErratumIn: PMID not supplied" | Builder rendering defect; the flag's content is correct (PubMed supplies no erratum PMID for PMID 25497547) and the card already says so. |
| AR-4 | Editorial | Final-score block, Nipbl paragraph, methods/provenance lines | 11 fused tokens separated: "Rao2014", "Step1"–"Step4", "day1", "day4", "Cutoff2026-10-08", "examinationUTC2026-10-09", "Consensus2Pro/0Deep/0API" | "randomize3808" rendering class, isolated rather than pervasive here. Gene/protein/complex names, DOIs and PMIDs left untouched. |
| AR-5 | Major | Consensus provenance block; `consensus-addendum.md` | "The earlier adversarial approval refers to its original edition" → "No earlier adversarial approval is attached to this case; the 2026-10-06 adversarial pass did not cover it" | False provenance inserted unconditionally by the 9 October revision script. 56 review files carry a 2026-10-06 approval banner and this case is not among them; the earlier edition carries no approval section. |
| AR-6 | Minor | Score card `final_score.confidence_rationale` | Access depth removed as a confidence determinant | The framework ties confidence to evidence density and reproducibility; the boilerplate rationale named "incomplete access" as a determinant. Rewritten to rest on density, replication and heterogeneity. Band re-tested against both band texts and unchanged. |

## Rejected findings (checked, no change)

1. **Level and aggregation.** Scope-matched is applied correctly: the gate is SC3 (broad primacy, Γ) because that is the subclaim whose scope matches the central claim, and SC2's Ζ is reported separately as the narrower locus-specific result rather than promoted to the headline. Δ rejected (positive tests are locus- and topology-specific, not scope-matching); Β rejected (native locus edits and distance-dependent activation supply real mechanistic content).
2. **`v5.claim_form` WARN ("phenomenon").** Rejected per convention: the restatement is a declarative causal proposition naming mechanism and consequence, and SC1 is explicitly held open as an existence statement whose empirical reality the Β entry does not dispute.
3. **Citation years and PubMed titles.** Refs [1] and [46] differ by one year (print/e-pub); PubMed's title fields for refs [29] and [60] drop the italicised *Shh* and *Sox2* that the review's reference list carries correctly. Both rejected per convention.
4. **Collective-author defect, empty `DOI: .` fields, retrieval handles.** None present.
5. **SC2's Ζ independence basis.** Checked against the shared-failure test: endogenous boundary editing, allele-resolved human/mouse developmental genetics, and locus-directed functional rescue are not defeated by one systematic error, and the review declines to count the repeated globin tethering reports as separate lines.
6. **Confidence band, Moderate versus High.** Tested per the lead's note. The High case is real (structure-collapse-with-expression-maintained is reproduced by at least four groups across four systems and three assay modalities); the Moderate case rests on concerns the review itself records (day-4 indirect effects, incomplete washout recovery, residual CTCF binding, reporter rather than endogenous contact-output measurement). Both are defensible on the framework text, so the framework does not force a change and the author's call stands. Flagged for the lead to settle corpus-wide.
7. **Revision regression.** None. The revised edition differs from the earlier one (`hsef_v53_popularity_ranks_41_45_2026-10-08/...`) only by the inserted provenance block. The specific regression reported elsewhere this round — a spliced Consensus block contradicting a preserved sentence denying Consensus work — is **absent here**: the preserved METHODS AND CONSENSUS CONTRIBUTION section already states Consensus was used through the signed-in browser, and REVIEWER NOTES already records "Consensus 2 Pro / 0 Deep / 0 API". `review_metadata.reviewer` reads "Codex" (house provenance) and was left unedited.

## Attribution audit

11 claims checked; all matched, including ones a looser reading would invert: all loop domains eliminated by cohesin loss while compartments and histone marks persist, affecting a minority of active genes (28985562); CTCF depletion removing TAD insulation while compartments stay segregated, with restoration instructive but transcriptome recovery incomplete (28525758); E–P interactions and transcription largely maintained after 3 h depletion, with cohesin loss reducing TF binding instead (36471071); TADs vanishing globally "even in the absence of transcriptional changes" with compartments reinforced (29094699); Fbn2 looped fraction ~3–6.5%, lifetime ~10–30 min (35420890); cohesin but not CTCF required for distant Shh activation and dispensable proximally (36097291); limb rewiring only when a CTCF boundary is disrupted (25959774); LDB1 DD4/5 dispensable for looping but essential for activation and FOG1/NuRD recruitment (24874989); the nonlinear contact–output relation measured in a deliberately simplified region (35418676); Sox2 bursting unassociated with enhancer proximity (31124784); ENCODE naming promoter class and enhancer–enhancer synergy beyond contact (42457959).

## Residual limitations

Several quoted figures (the 3 h transcript counts; Rao's 87% under a 30% criterion; Nora's 18% residual boundaries and 370→4,996 genes; Nipbl's 637/487/1,107; ENCODE's 92 million links and 10,356 benchmark pairs) are Methods/Results-level and confirmable here only against the review's own partial-body receipts. Access depth unchanged (11 partial-body, 49 complete abstract/status). Validator FAIL is a framework-version mismatch; the WARN is the rejected claim-form false positive.

## Strings for the lead to propagate to package-level files

| Old | New |
|---|---|
| `Decisive work underway` (3d-genome-regulation row, Momentum column) | `Mechanistic only` |
| `The earlier adversarial approval refers to its original edition; this provenance supplement is not a new independent or blinded adjudication.` | `No earlier adversarial approval is attached to this case; the 2026-10-06 adversarial pass did not cover it, and this provenance supplement is not a new independent or blinded adjudication.` |

No level, confidence, role, source count or claim count changed. The earlier edition at `hsef_v53_popularity_ranks_41_45_2026-10-08/3d-genome-regulation/` is a divergent copy and was **not** overwritten.

**Lead policy recorded (confidence bands, 2026-10-14).** The lead adjudicated the question raised in this pass and set a two-step test for the corpus. Step 1: strike access depth as a determinant everywhere — how much of a paper the reviewer could read is neither evidence density nor reproducibility, and it never justifies a band. Step 2: re-derive the band from what remains; if access depth was the only thing holding the band down and the cited evidence meets the High text, raise it, but if the review documents substantive evidence-quality or reproducibility concerns (indirect effects, incomplete washout, residual confounding of the mechanism, reporter rather than endogenous readouts) those are valid determinants and the band stands. A reviewer's uncertainty about the evidence is a legitimate reason for Moderate; a reviewer's uncertainty about their own reading access is not. This pass's rationale rewrite is step 1; the band decision below is step 2. Step 1 applied (rationale rewritten under AR-6). Step 2: band stands at Moderate, and the lead has confirmed this call. The reservations the review records in its own text — day-4 indirect effects, incomplete transcriptome recovery after washout, residual CTCF binding at some boundaries, and reporter rather than endogenous measurement of the contact-output relation — are valid determinants under step 2.

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