CORRECTED CONSOLIDATED ODD-PRIME PUBLICATION PACKAGE
Scott Schubr -- October 18, 2026 -- correction revision 1

BASELINE AND RELATIONSHIP
This is a separately authorized revision of the consolidated odd-prime assessment
package. Its sealed original and the earlier qutrit original remain unchanged:
 D:/LUMINARY ODDITIES LLC/AI Research/odd_prime_learner_prepublication_Riemann_A_20261018
 ZIP SHA256 35e4dffd9a95256ba90cb957dc6e261e53aa99e9086ee3e8cea7ad7b051a18ef
 D:/LUMINARY ODDITIES LLC/AI Research/mixed_qutrit_learner_prepublication_Riemann_A_20261018
 ZIP SHA256 2afda27faa4b26a86e96cd04da481075c093109115c67d0a498b246c62589e75
The qutrit result is the p=3 case. This is ONE consolidated paper in the same
result family, not two overlapping independent publications.

CHANGES AND EFFECT
1. Added version-pinned Iyer-Liang, Holt-Subramanian and Bittel et al. references
   and precise task comparisons. The earlier mixed-qubit positive Fourier-dual
   overlap distribution and general difference sampling are credited. Exact mixed
   stabilizer-group learning, high-overlap mixed-qubit single-copy learning, and
   exact pure prime-dimensional learning are distinguished from this theorem.
2. Replaced numerical assertion acceptance gates with explicit exceptions that
   remain active under optimized Python. No numerical formula, tolerance, seed,
   tested state, or mathematical acceptance threshold changed.
3. Hardened the manifest checker: explicit unique entries and JSON keys, strict
   schema/SHA format, portable canonical paths, Windows unsafe path rejection,
   and resolved containment before opening payloads; exact file-set checks remain.
4. Clarified the auxiliary favorable witness: absorb on first recorded optimizer,
   kill at the first unfavorable continuation or abort, and require conditional
   states only on active favorable prefixes. This is proof exposition; the
   physical algorithm, conditioning budgets and success bound are unchanged.
5. Changed the assessment-draft label to a corrected publication version and
   updated development disclosure for author publication. Set the bibliography
   at legible 10pt so all ten references fit on the final ninth page. The four
   algorithm steps now appear together on page 7.

UNCHANGED MATHEMATICS
Both theorem statements, domains, constants, resource formulas and algorithms
are unchanged. proof/GENERAL_PROOF.txt retains its original SHA256
c041807f92005436abdfae2c10ec583f17453903c4acbc7d7ec0d5d2763b9ef8.
The manuscript contains the complete analytic proof. review/manuscript.diff
records the exact source changes against the sealed all-prime original.

EVIDENCE ORGANIZATION
The baseline/ directories contain explicitly historical producer/Publisher records
and earlier outputs, bound to their earlier sources. Their old permissions,
page counts and hashes describe that historical edition. They are not fresh
reviews of this revision. Current source-bound review and replay receipts are at
review/ and verification/ top level. PROVENANCE.json binds current sources.
MANIFEST.sha256.json binds the final complete package, excluding itself.
Originals-before/after identities and ZIP readback establish preservation,
not mathematical correctness.

AUTHORIZATION AND LIMITS
Scott directly requested the minor revisions, preservation of both originals,
fresh verification and delivery to Luminary Research Archive for publication.
The corrected consolidated version is the authorized publication candidate.
Delivery does not itself establish observed publication or public acceptance.
Analytic validity and provisional specialist novelty/value were independently
reviewed; assistant reviews are not external human refereeing. Finite complex128
diagnostics corroborate the proof; they do not prove all n, simulate the full
learner, establish optimal costs, or certify worldwide priority.
