Fix calibration scaling and complete the bounded release-preparation pass

.github/repository_metadata.json

@@ -0,0 +1,5 @@
+{
+  "description": "Quantitative moonshine VOA identification from near-extremal correlations, with exact certificates and a physics-first CFT tutorial.",
+  "topics": ["conformal-field-theory", "vertex-operator-algebras", "moonshine", "monster-group", "correlation-functions", "rigidity", "interval-arithmetic", "mathematical-physics"],
+  "scope": "Desired About-sidebar values; this file does not itself change GitHub repository settings."
+}

.github/workflows/quantitative-core-review.yml

@@ -7,7 +7,7 @@ on:
 permissions:
   contents: read
 jobs:
-  verify:
+  preserved_history:
     runs-on: ubuntu-24.04
     timeout-minutes: 5
     env:
@@ -38,6 +38,7 @@ jobs:
           persist-credentials: false
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
         with:
+          ref: ecd0623e9de41b49754b6ca16dc23752443143af
           path: candidate
           persist-credentials: false
       - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
@@ -92,7 +93,7 @@ jobs:
           cmp /tmp/pr31-normal.json /tmp/pr31-O.json
           cmp /tmp/pr31-normal.json /tmp/pr31-OO.json
           cat /tmp/pr31-normal.json
-      - name: Current fenced-math checks and exact presentation delta
+      - name: Frozen pre-release fenced-math and presentation evidence
         run: |
           cd candidate
           python3 checks/verify_teaching.py --baseline ../displaybase > /tmp/display-normal.json
@@ -102,7 +103,7 @@ jobs:
           cmp /tmp/display-normal.json /tmp/display-OO.json
           cat /tmp/display-normal.json
           python3 checks/verify_display.py --export-math /tmp/monster-math.json
-      - name: Parse every active-page expression with MathJax
+      - name: Parse the frozen pre-release expression corpus
         run: |
           npm install --prefix "$RUNNER_TEMP/monster-math" --ignore-scripts --no-audit --no-fund mathjax-full@3.2.1
           cd candidate
@@ -151,3 +152,45 @@ jobs:
           PYCODE
           git diff --check
           test -z "$(git status --porcelain)"
+  current:
+    runs-on: ubuntu-24.04
+    timeout-minutes: 8
+    env:
+      PYTHONDONTWRITEBYTECODE: '1'
+      OPENBLAS_NUM_THREADS: '1'
+      OMP_NUM_THREADS: '1'
+      MKL_NUM_THREADS: '1'
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          ref: ecd0623e9de41b49754b6ca16dc23752443143af
+          path: releasebase
+          persist-credentials: false
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+        with:
+          path: current
+          persist-credentials: false
+      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
+        with:
+          python-version: '3.13.5'
+      - name: Install the pinned reproduction dependency
+        run: python -m pip install -r current/requirements-reproduce.txt
+      - name: Current full suite, exact delta, and separate historical outcomes
+        run: |
+          cd current
+          python verify_current.py --all --baseline ../releasebase --output "$RUNNER_TEMP/current-reports"
+          python checks/verify_display.py --export-math "$RUNNER_TEMP/current-math.json"
+          git diff --check
+          test -z "$(git status --porcelain)"
+      - name: Parse the complete current reader corpus
+        run: |
+          npm install --prefix "$RUNNER_TEMP/current-renderer" --ignore-scripts --no-audit --no-fund mathjax-full@3.2.1
+          NODE_PATH="$RUNNER_TEMP/current-renderer/node_modules" node current/checks/render_math.cjs "$RUNNER_TEMP/current-math.json"
+      - name: Retain current verification stdout and stderr
+        if: always()
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
+        with:
+          name: current-verification-reports
+          path: ${{ runner.temp }}/current-reports
+          retention-days: 30
+          if-no-files-found: warn

CHANGELOG.md

@@ -0,0 +1,30 @@
+# Changelog
+
+## Unreleased — release preparation, 27 September 2026
+
+Corrected a small-scale interval-calibration exception without changing the
+analytic theorem, exact premises, or sufficient error tolerances. Added adaptive
+exact square-root enclosures, an explicit precision-work-limit outcome, and
+scale/interval regression checks. Current report fingerprints are separate from
+preserved historical outputs.
+
+Added `verify_current.py`, a pinned reproduction requirements file, a
+reproduction guide, current release integrity checks, and a clear distinction
+between exact certification, historical strict replay, and bounded portability.
+
+Clarified current versus historical status, indexed preserved and pending work,
+and replaced ambiguous missing-file references with pinned destinations. Added
+citation metadata, a scope-accurate LLM source map, and release-scope notes.
+The Gaberdiel-first physics teaching path and approved native mathematics remain.
+
+No manuscript, DOI, tagged release, or merge of unrelated research branches is
+included. Protected historical notes, provenance archives, the original license,
+and existing recorded evidence are retained.
+
+## Prior integrated research and presentation
+
+The selected quantitative core, calibration/sharpness results, and bounded
+review were integrated before this release-preparation pass. Five physics
+lessons and subsequent math-rendering improvements form the current teaching
+path. Consult the [archive index](docs/ARCHIVE.md) and Git history for exact
+historical commits rather than interpreting those integrations as formal releases.

CITATION.cff

@@ -0,0 +1,24 @@
+cff-version: 1.2.0
+message: "Please cite this research repository and include the exact commit or tagged version used. Citation is a scholarly request, not an additional license condition."
+type: software
+title: "Monster Correlation Rigidity"
+authors:
+  - family-names: "Lin"
+    given-names: "Ruge"
+    email: "gogoko699@gmail.com"
+repository-code: "https://github.com/GoGoKo699/Monster-Correlation-Rigidity"
+url: "https://github.com/GoGoKo699/Monster-Correlation-Rigidity"
+license: MIT
+abstract: >-
+  Research proofs, exact finite certificates, and a physics-first tutorial for
+  quantitative conditional identification of the moonshine vertex operator
+  algebra from near-extremal correlation coefficients. The exact ambient
+  assumptions and imported classification endpoint remain explicit.
+keywords:
+  - conformal-field-theory
+  - vertex-operator-algebras
+  - moonshine
+  - monster-group
+  - correlation-functions
+  - rigidity
+  - interval-arithmetic

PHYSICS_STATUS.md

@@ -1,3 +1,5 @@
+> **Historical record, not the current project status.** The dated ledger below is preserved unchanged. Start with [current scientific status](docs/RESEARCH_STATUS.md), [the physics learning path](docs/learn/README.md), or [the archive index](docs/ARCHIVE.md).
+
 # Physical-selection ledger — 25 September 2026
 
 Latest base: `ba02084a1a74fc0b4db1299f6ab3bc327a0901ef`. Physical selection remains the objective. STATUS_CURRENT.md and STATUS.md retain the former certification and import records unchanged.

README.md

@@ -47,19 +47,52 @@ Finite coefficient tolerances are not approximate VOA axioms or a device-indepen
 
 ## Reproduce the selected certificates
 
+The current exact checks and documentation checks use Python's standard library:
+
 ```sh
-python checks/verify_extraction_core.py
-python checks/verify_calibration_and_sharpness.py
-python checks/audit_quantitative_core.py
-# Each script also runs unchanged with python -O and python -OO.
+python verify_current.py
 ```
 
-The preserved core reports have 63 and 74 labeled checks. The final review adds 23 labeled controls, including exact interval and independent rational-oracle cases. No full Monster tensor is simulated. Passing finite checks is not a proof-assistant verification of the source theorems or an independent human review.
+For the complete checker inventory and the separate historical comparisons,
+use the [pinned reproduction environment](docs/REPRODUCIBILITY.md):
 
-The historical `verify.py` and `checks/replay_portability.py` are unchanged. Their strict replay and bounded-portability outcomes are reported separately; a green overall job is not a claim of byte-identical historical floating-point output.
+```sh
+python -m pip install -r requirements-reproduce.txt
+python verify_current.py --all
+```
 
-## Earlier research and preservation
+The wrapper runs each selected checker in normal, `-O`, and `-OO` modes and
+compares the current exact report fingerprints. The original 63-, 74-, and
+23-check evidence is retained at its recorded versions. The repaired calibration
+has its own current fingerprint and additional scale-regression checks; see
+[calibration precision](docs/CALIBRATION_PRECISION.md) and
+[current report metadata](results/current_reports.json).
 
-The earlier [physical-selection ledger](PHYSICS_STATUS.md), [claim ledger](STATUS.md), numbered research notes, immutable archives, code and recorded results remain available as historical research. Their older “current” labels do not supersede [the active work order](work_orders/CURRENT.md) or [the present status](docs/RESEARCH_STATUS.md). Unrelated pending branches are not merged merely to create this reader path.
+The historical `verify.py` is **not** the current suite. Its strict byte replay
+and the unchanged bounded-portability comparison are reported separately.
+A historical floating-point mismatch remains a strict failure even when the
+current exact suite and bounded comparison pass. The reproduction guide explains
+exit codes and the optional strict-replay requirement.
+
+No full Monster tensor is simulated. Finite checks do not verify the imported
+classification theorems or constitute independent human review.
+
+## Earlier research and preservation
 
-Original code and documentation retain the owner's [MIT license](LICENSE). Imported mathematical results remain attributed to their authors. Research inquiries and corrections may be sent to Ruge Lin at gogoko699@gmail.com.
+Use [the archive and pending-work index](docs/ARCHIVE.md) for older gate-testing,
+circuit, and readout records. The dated ledgers now carry historical banners.
+Their older “current” labels do not supersede
+[the current scientific status](docs/RESEARCH_STATUS.md). Protected source notes,
+provenance archives, and saved evidence remain available unchanged; legacy
+mathematics may not render correctly in every GitHub viewer.
+
+[Release scope](docs/RELEASE.md), [the changelog](CHANGELOG.md), and
+[citation metadata](CITATION.cff) describe the current research package without
+claiming a manuscript, publication, DOI, or tagged release. For LLM-assisted
+research, [the source map](llms.txt) points questions to the theorem, its exact
+assumptions, teaching anchors, verification, and historical boundaries.
+
+Original code and documentation retain the owner's [MIT license](LICENSE).
+Imported mathematical results remain attributed to their authors. **Manuscript
+writing remains on hold.** Potential collaborators, readers with corrections,
+and other research inquiries may contact Ruge Lin at gogoko699@gmail.com.

STATUS_CURRENT.md

@@ -1,3 +1,5 @@
+> **Historical record, not the current project status.** The dated ledger below is preserved unchanged. Start with [current scientific status](docs/RESEARCH_STATUS.md), [the physics learning path](docs/learn/README.md), or [the archive index](docs/ARCHIVE.md).
+
 # Current claim ledger — 25 September 2026
 
 Latest continuation base: `85215c45005a3241485b99b57b2dc9de63b85702`.

audits/calibration_scale_erratum.md

@@ -0,0 +1,52 @@
+# Calibration implementation erratum: positive small denominators
+
+Baseline: `ecd0623e9de41b49754b6ca16dc23752443143af`.
+Affected code: `checks/verify_calibration_and_sharpness.py`.
+The normalized-correlation theorem, calibration identities, and all physical
+hypotheses and comparison tolerances are unchanged.
+
+## Reproduced defect
+
+For exact orthogonal Ising stress tensors $e,f$, let $w_1=s e$, $w_2=s f$ with
+$s=2^{-39}$. The seven exact scalar inputs in the public function's order are
+
+```math
+(s^2/4,\ s^2/4,\ s/4,\ s/4,\ s^3/2,\ s^3/2,\ 0).
+```
+
+Each projected norm is $(47/192)s^2$. The positive cross denominator equals
+$(47/48)2^{-80}$, but the old absolute $2^{-80}$ grid gives it a zero lower
+endpoint. The subsequent interval division raises `ValueError`. This is an
+implementation exception on valid scaled inputs, not a false-positive
+certificate or a counterexample to the analytic theorem.
+
+The [pinned audit report](https://github.com/GoGoKo699/Monster-Correlation-Rigidity/blob/62cba790ac69b8cf7ad339ad8d0aa21578366770/audits/release_sanity_20260927.md)
+and its recorded runner logs preserve the original finding.
+
+## Correction and evidence
+
+`checks/scale_safe_roots.py` refines the dyadic grid using exact numerator and
+denominator bit lengths. The [arithmetic proof](../docs/CALIBRATION_PRECISION.md)
+shows that every positive accepted radicand receives a positive lower root bound
+and an outward enclosure. The public calibration function returns an explicit
+inconclusive result at the documented square-root work limit.
+
+`checks/verify_calibration_scales.py` reproduces the old zero-lower-endpoint
+calculation without restoring the old bug. It tests the corrected public path
+on exact and finite-width independently scaled pairs, very small/large scales,
+known stress shifts, near-zero norm boundaries, negative signs, duplicate fields,
+impossible couplings, and the resource-limit case. Existing calibration controls
+and the adversarial rational/bisection audit are also rerun on the repaired code.
+
+The old 74-check stdout is not overwritten. The refined enclosures change that
+output; [current report metadata](../results/current_reports.json) records the
+new fingerprint separately. The earlier audit's source-hash expectation is
+updated explicitly for the repaired source, while its mathematical and
+implementation controls are retained. All historical source bytes remain
+accessible at the baseline commit, and all pre-existing saved report files
+remain unchanged.
+
+No Monte Carlo or full VOA simulation is used. A false or inconclusive answer
+does not establish that the theory differs from moonshine. The certificate
+continues to assume the exact ambient class, weight, real structure, and known
+stress tensor; it does not infer those premises from the seven numbers.

audits/release_preparation.md

@@ -0,0 +1,77 @@
+# Release-preparation integration
+
+Base commit: `ecd0623e9de41b49754b6ca16dc23752443143af`.
+Base tree: `20c5252c0bd5e83d6a908ee226e5799e947280fc`.
+The repository owner authorized the bounded repairs from the pre-release audit
+and modifications to the GitHub-facing documentation. No release, manuscript,
+DOI, outreach, or unrelated research-branch merge is included.
+
+## Source and preservation
+
+A read-only GitHub Actions artifact supplied the exact baseline source archive.
+Its 110 extracted files reproduced the baseline Git tree above. Local tests
+therefore use the actual complete source bytes, not a hand-reconstructed subset.
+Network Git cloning in the local container was unavailable.
+
+The release manifest declares every source change and addition and fingerprints
+all other baseline files. Protected numbered research notes, provenance files,
+original license, historical root verifier, bounded-portability policy, and all
+pre-existing saved result files remain byte-identical. The gate-testing and
+physical-selection ledgers retain their bodies below new historical banners.
+The seed-protected `STATUS.md` remains wholly unchanged; the archive index
+identifies it as historical. Canonical proof mathematics is
+compared expression by expression; only provenance and implementation pointers
+were added to those proofs.
+
+## Calibration correction
+
+The small-scale exception is fixed by scale-aware exact square-root enclosures.
+The public function catches a typed precision-work-limit condition and returns
+an explicit inconclusive result. The seven-scalar identities, physical premises,
+and numerical comparison tolerances are unchanged. The original calibration
+controls and independent bisection audit run on the corrected implementation.
+The new scale checker covers exact and finite-width independently scaled fields,
+stress shifts, tiny and large scales, sign/cap/duplicate controls, and the
+resource-limit case. See [the erratum](calibration_scale_erratum.md).
+
+Current exact outputs are recorded separately in
+[current report metadata](../results/current_reports.json). Old 74-check and
+23-check report fingerprints remain as historical evidence, not falsely
+relabelled as the current implementation's output. The extraction output is
+unchanged. Normal, `-O`, and `-OO` outputs agree in the local exact-check replay.
+
+## Reader and reproduction repair
+
+The current wrapper separates selected exact checks, documentation checks,
+historical strict replay, and bounded portability. A pinned reproduction file,
+archive/pending-work index, citation metadata, source map, changelog, and release
+scope accompany it. The old integration workflow still executes its complete
+chain on the frozen pre-release tree; an additional job checks the real candidate
+against that tree and runs the complete current suite. This is not a replacement
+of the archived science by a formatting-only check.
+
+The desired About-sidebar description/topics are stored in
+[repository metadata](../.github/repository_metadata.json). File creation does
+not itself alter GitHub repository settings; no settings-write result is implied.
+
+## Verification boundaries
+
+The exact candidate's CI stdout/stderr is retained as a workflow artifact.
+The pull-request validation record contains the actual final runner results,
+including strict-versus-portability outcomes; an aggregate green workflow is not
+a claim of strict byte identity. Local rendering uses Chromium/MathJax previews,
+not GitHub's live browser. The active page inventory is explicit in the release
+manifest; legacy rendering limitations remain documented in the archive index.
+
+These are implementation, preservation, and release-readiness checks, not a new
+scientific campaign, independent human review, complete audit of the research
+backlog, or proof-assistant verification. Manuscript writing remains on hold.
+
+## Integration regression caught before merge
+
+The first candidate added a historical banner to the seed-protected `STATUS.md`.
+The complete CI suite correctly rejected that edit through the unchanged
+provenance verifier. The banner was removed and the exact original file restored;
+`STATUS.md` is now also explicitly protected by the current integrity check.
+No historical checker, source hash, or numerical tolerance was weakened to
+accept the edit. Current readers are directed through the archive index instead.

checks/audit_quantitative_core.py

@@ -193,13 +193,14 @@ def calibration_audit():
 
 def main():
     files={'verify_extraction_core.py':'d2062b1577beb7a34c0665fc358da7b1daf846886cd48dfa0b26acf83c107e83',
-           'verify_calibration_and_sharpness.py':'d2c16a6c9c01af9d1198cfc2b4459d78966a42c314b8a0575bd37626bcd2eef1'}
+           'verify_calibration_and_sharpness.py':'7a2b98dab6df116ea804706d75e8fb5a0d246ea889f6381f13e88dd85bedbb5a'}
     for path,h in files.items():
         need(hashlib.sha256(Path(__file__).with_name(path).read_bytes()).hexdigest()==h,
              'preservation:'+path)
     structural_identities(); interval_audit(); calibration_audit()
     print(json.dumps({'status':'PASS bounded adversarial implementation and normalization checks',
-        'reviewed_head':'6d460bef7041f17cba1658c99d6797ef4b00275c',
+        'original_reviewed_head':'6d460bef7041f17cba1658c99d6797ef4b00275c',
+        'implementation_scope':'scale-safe calibration repair; original controls retained',
         'checks':len(LABELS),'labels':LABELS,'case_counts':COUNTS,
         'arithmetic':'rational arithmetic; independent bisection oracle',
         'largest_physical_coefficient_vector':3,'full_VOA_simulated':False,