Skip to content

P104: portable 233-CLEAR conditional κ ≥ 0.0037 - #7637

Open
SYLforge wants to merge 1 commit into
CrocSwap:mainfrom
SYLforge:research/final233-portable-20261118
Open

SYLforge wants to merge 1 commit into
CrocSwap:mainfrom
SYLforge:research/final233-portable-20261118

Conversation

@SYLforge

Copy link
Copy Markdown

Summary

Portable 233-CLEAR composition supporting conditional κ ≥ 0.0037, combining the frame-restored H5B construction with the cited E1/E4 follow-up changes. This is a research submission, not a selected-frontier update or an unconditional multiplication theorem.

  • 78,596 events; 233 nonzero CLEARs; peak live stock 50; 52 used labels.
  • All 68 reserved labels and 1,156 signed banks remain charged.
  • Changed entrance: initial[457]: 3 -> 281.
  • Canonical candidate SHA-256: 8ab3d9dbe2e2e24202ac96c6122b010b3e7980912f935cc0cc2db2e9d85f83aa.

Reproducible package

Adds 24 files under research/final233-portable-conditional/. Python sources remain visible; large immutable inputs and retained evidence are deduplicated in data/inputs.tar.gz, with original paths and exact SHA-256 values in DATA-MANIFEST.json. Operational logs and duplicate copies are omitted.

See README.md for Python 3.12 commands, dependencies and the pinned external #7334 archive. The archive must be downloaded or supplied locally with its documented hash; this is an explicit source-availability dependency. Integrity without it reports EXTERNAL_NOT_CHECKED. Materialization/reconstruction uses a fresh work directory.

Validation and precise scope

  • Compact reconstruction passed and reproduced the canonical candidate hash (78,596 events). Local package integrity passed.
  • Earlier candidate-adapted native execution passed (exit 0, 432.298 s).
  • Earlier signed-integrated execution passed (exit 0, 191.156 s), covering 3,060 catalogue maps: 2,000 guarded authenticated cached results and 1,060 fresh executions.
  • Earlier independently authored finite audit passed: 233 recoverability identities and 932 targeted mutation predicates.

The native, signed and finite-audit results above are retained prior author evidence, not new scientific executions of this compact repackaging. See VALIDATION.md for the exact chronology and limits. No independent human or formal-kernel review, physical-provider admission, or selected-frontier acceptance is claimed.

Conditions and verifier distinction

The five inherited framework assumptions remain. A concrete common-U supplier satisfying the changed entrance, whole-private-extent/nonaliasing, reset, residual, precision and charged-cost obligations has not been established. See CONDITIONS.md. The finite scalar audit is not a physical reset proof.

The candidate-specific adapters are distinguished from the original project verifiers. This PR does not modify shared project verifiers or install an acceptance gate for other researchers. The candidate hash identifies these exact candidate bytes; it is not a restriction on alternative candidates or verification approaches.

Attribution and continuation

Credits: Ali Romman/#7605, SYLforge/H5B, DreamingOfClouds/#7630, Clear1nsight/#7631/#7633 and #7626, Rohan Arun, and the retained upstream contributors. LICENSE, NOTICE, PROVENANCE.md and SOURCES.json preserve lineage. Substantial OpenAI assistance was used.

The next review step is independent reconstruction followed by examination of the changed entrance and a concrete supplier's whole-extent/reset/tariff contract. The open blocker is the absence of an admitted supplier, not a requirement that other research pass our candidate-specific hash check.

Add compact 24-file reproducible research package with preserved inputs, provenance and explicitly conditional validation scope.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant