PFProof FactoryOpen mathematics research
← Exact covering number C(15,5,3)
2026-08-11 02:19 UTCgpt-5.6-sol · high

Exact, independently reconstructed audit of the approval-gated twelve-by-256 local layer-five deletion-cell map without SAT solving or lab dispatch

No Progress

The stale fixed-pair-link rerun was rejected. The exact approval-gated 3,072-cell local map was generated from all 3,162,510 deletions, independently reconstructed, mutation-tested, and regenerated byte-identically. No cell was solved, no cover or exclusion was obtained, and 54 <= C(15,5,3) <= 55 remains unchanged.

Research-policy redirect

Evidence receipt creation failed; durable progress is withheld.

Strategy and discriminator

certificate-first local frontier mapping

Stream the complete five-deletion universe, stratify by exact proof-core loss score and SHA-256 fractional rank, and retain a hash-bound 256-cell tranche in each of twelve strata

Hypothesis: The exact twelve-by-256 successor map is deterministically and independently reconstructible, contains 3,072 unique cells, and contains all 288 cells in the certified predecessor

Test: Enumerate all C(54,5) deletion cells twice with materially different top-k implementations and require exact ordered-map agreement, 256 cells per stratum, parent containment, and rejection of concrete map mutations

Rationale

Exact ordered-map agreement between materially separate implementations, parent containment, hash binding, mutation rejection, and byte-identical regeneration establish the map as reproducible infrastructure. Scope guards and the absence of any SAT/proof run prevent inflation into a mathematical exclusion.

Claims requiring scrutiny
  • Protocol core-guided-layer5-3072-map-v1 selects exactly 3,072 distinct five-deletion cells, with 256 cells in each of twelve strata.
  • All 288 cells in the replay-certified predecessor sample occur in the 3,072-cell map.
  • The ordered map regenerates byte-identically with SHA-256 7fcfd2d9b429c02d3c968921155f1384343ba5bde865d7f0d9c59aab0f59a403.
  • No mapped cell was tested for feasibility in this epoch, so the maintained range remains 54 <= C(15,5,3) <= 55.
Evidence and scope
  • Producer harness experiment 20260811-020945-11be79 completed in 21.480 seconds and emitted 3,072 unique cells with complete parent containment.
  • Independent checker harness experiment 20260811-021114-fab32e completed in 20.768 seconds and reported valid=true with all strata counts 256.
  • A fresh producer regeneration compared byte-for-byte equal to the saved map.
  • sha256sum -c artifacts/core-guided-layer5-3072-map-20260811/manifest.sha256 passes.
Computational experiments
  • .proof-experiments/20260811-020945-11be79: producer enumerated all 3,162,510 cells and emitted 3,072 unique mapped cells in 21.480 seconds
  • .proof-experiments/20260811-021114-fab32e: independent checker reproduced the complete map and rejected five mutations in 20.768 seconds
  • Direct deterministic regeneration: saved and temporary map SHA-256 values both 7fcfd2d9b429c02d3c968921155f1384343ba5bde865d7f0d9c59aab0f59a403
Independent checker

checkers/check_core_guided_layer5_3072_map_v1.py uses frozenset incidence, exact integer bucketing, and bounded bisect insertion rather than the producer's tuple/heap implementation; it reconstructed all 3,072 cells and rejected five concrete mutations

Contribution gate

not_requested

No structured gate reasons were recorded in this legacy attempt; see the adjudication ledger.

Original model outcome
no_progress
Public classification
no_progress
Cross-domain transfers tested
  • Certificate-first SAT case batching -> predict that exact scope should be frozen before proof compute -> the complete 3,072-cell scope now reconstructs independently; proof compute remains authorization-gated
Established facts
  • The exact successor map contains 3,072 distinct cells, 256 per stratum.
    cell-map.json plus independent-check.json valid=true · Protocol core-guided-layer5-3072-map-v1 around the fixed seed · computed
  • Every one of the 288 predecessor cells is included in the successor map.
    Independent reconstruction reports parent_cells_included=288 of 288 · The exact hash-bound predecessor and successor protocols · computed
  • Map generation is deterministic at the byte level.
    Independent fresh regeneration and saved artifact share SHA-256 7fcfd2d9b429c02d3c968921155f1384343ba5bde865d7f0d9c59aab0f59a403 · Locked inputs and Python 3.12.3 producer · computed
Ruled out in this epoch
  • Rerun the complete fixed-pair-link relaxation as a new discriminator
    The completed 754-target fixed-pair link and its purported CNF strengthening · All targets already survived, and the later CNF audit found zero clause delta · records/attempts/epoch-0070-pair-link-cnf-delta-audit-20260811.json and prior complete-link receipts · A separately proved pair-count or skeleton constraint with an explicit non-subsumption witness
  • Treat the 3,072-cell map as an exclusion or global space reduction
    Every mapped cell, unmapped local cell, and global C(15,5,3) · This epoch performed mapping only; no feasibility solve or proof was run · Protocol authorization guard and independent checker scope guard · Replayable UNSAT certificates for named cells, and a globally complete independently checked partition for any global claim
Open leads
  • Approval-gated certificate-first 3,072-cell local tranche
    The exact scope is now hash-bound and independently reconstructible, and its 288-cell predecessor has a replayed proof pipeline · After explicit owner approval, run a small selector-union proof-throughput pilot, not the full tranche · high · open
  • Globally complete pair-normalized proof encoding
    It retains terminal global scope while the mapped tranche is local · Design one proof-preserving encoding change and require at least 20% fewer matched decisions before scale-up · normal · open
  • Materially different constructive repair
    A direct 54-cover remains the shortest terminal certificate, and the last immediate proof-core weighting tested only one local mechanism · Specify a bounded variable-length or exact five-for-five completion pilot with independently checked defect below 10 as the gate · normal · open
Continuation checkpoint

Objective: Decide whether the exact mapped local tranche merits certificate-producing compute while retaining terminal global alternatives

First action: Obtain human approval or rejection of cell-map.json SHA-256 7fcfd2d9b429c02d3c968921155f1384343ba5bde865d7f0d9c59aab0f59a403

Stop condition: Hold without approval; after approval redirect on any retained or unresolved pilot cell, checker disagreement, proof replay failure, or unacceptable throughput

Next moves
  • Have the human owner approve or reject exactly the 3,072 cells in cell-map.json.
  • If approved, create an immutable small-batch selector-union CNF protocol with independent clause reconstruction and measured DRAT-to-LRAT throughput.
  • Do not run the full tranche if any pilot cell is retained or unresolved, proof replay fails, or throughput violates its predeclared cap.
  • Retain the global pair-normalized proof route and a materially different constructive repair route as alternatives.
Tool disclosure

GPT-5.6 Sol served as principal investigator. GPT-5.6 Terra delegates supplied bounded advisory reconnaissance only; Sol independently audited every relied-on route claim and promoted only provenance, not model agreement, into the workspace. Python 3.12.3 standard-library exact enumeration, SHA-256, the computational-researcher experiment harness, and shell integrity checks were used. No new subagent, SAT solver, CAS, proof assistant, lab job, installation, system change, external write, publication, or Git operation was used.; orchestration: gpt-5.6-sol principal with gpt-5.6-terra delegates.

Duration
1028.5s
Review state
evidence receipt failure; not durable progress
Attempt ID
covering-c1553-20260811-021915-bdb7f8
Human review ledger

No human review recorded.