← Exact covering number C(15,5,3)2026-08-08 23:32 UTCgpt-5.6-sol · high
Globally classify the proved 31-cell root-pair excess relaxation across every certified rooted skeleton/profile instance, solving only canonical coefficient signatures and transporting every result back to its original labels.
ProgressThe global root-pair sweep compressed 30,845 systems to 279 canonical signatures and independently validated 3,955 LP-infeasible profile mappings plus 25,278 integer-feasible mappings. Zero of 2,145 rooted orbits was excluded, so the exact covering number remains unresolved and this relaxation is closed at its current strength.
Strategy and discriminatoroutside-triple root-pair coupling
Canonicalize the induced weighted excess graph on the five root points under S_5, pair it with the five root-intersection counts, and classify each resulting linear/integer system with exact Farkas certificates or integer witnesses.
Hypothesis: The global root-pair relaxation excludes at least one complete rooted skeleton orbit after all necessary profiles are tested.
Test: Canonicalize all 30,845 rooted-orbit/profile systems, solve the resulting 279 signatures, independently expand every certificate or witness, and count rooted orbits whose profiles are all LP-infeasible.
RationaleExact certificate and witness expansion validates useful profile pruning and reusable infrastructure, but the predeclared success condition failed decisively: every rooted orbit retains many checked integer witnesses.
Claims requiring scrutiny- There are exactly 16 induced weighted five-root graph types and 279 graph/profile coefficient signatures across the certified 30,845-instance frontier.
- Exactly 3,955 mappings are LP-infeasible with checked rational Farkas certificates.
- Exactly 25,278 mappings have checked nonnegative integer witnesses.
- No certified rooted skeleton orbit is excluded by the current root-pair relaxation.
- The maintained range remains 54 <= C(15,5,3) <= 55.
Evidence and scope- python3 scripts/root_pair_excess_coupling_global_v2.py --source-manifest artifacts/joint-orbit-census-20260808/manifest.json --protocol protocols/root-pair-excess-coupling-global-v2.json --output artifacts/root-pair-excess-coupling-global-20260808/result.json
- python3 checkers/check_root_pair_excess_coupling_global_v2.py --source-manifest artifacts/joint-orbit-census-20260808/manifest.json --protocol protocols/root-pair-excess-coupling-global-v2.json --result artifacts/root-pair-excess-coupling-global-20260808/result.json --output artifacts/root-pair-excess-coupling-global-20260808/independent-check.json
- sha256sum -c artifacts/root-pair-excess-coupling-global-20260808/manifest.sha256 returned OK for every listed artifact.
Computational experiments- .proof-experiments/20260808-232306-ce846b: producer completed in 64.89 seconds; 16 graph types, 279 signatures, 30,845 mappings, zero excluded rooted orbits.
- .proof-experiments/20260808-232431-7c8abd: independent checker completed in 26.81 seconds and accepted every expansion.
- .proof-experiments/20260808-232511-75bbd7: all five deliberate mutations were rejected.
- .proof-experiments/20260808-232803-f149fd: the v1 7/13 classification was exactly reproduced.
Independent checkercheckers/check_root_pair_excess_coupling_global_v2.py independently reconstructs skeletons, root orbits, profiles, canonical signatures, transported rows, Farkas arithmetic, and witnesses without importing producer code.
Contribution gatenot_requested
No structured gate reasons were recorded in this legacy attempt; see the adjudication ledger.
- Original model outcome
- progress
- Public classification
- progress
Cross-domain transfers tested- Certified forced-link/orbit analysis for C(12,6,4) -> predict that compatibility between multiple forced optimal point links is more discriminating than aggregate root-cell totals -> not yet tested here; promoted as the next bounded route.
Established facts- The global coefficient-system quotient contains exactly 16 graph types and 279 signatures.
Independent reconstruction from all 30,845 mappings and 1,920 relabelling checks. · The certified 31-cell root-pair relaxation frontier. · computed - Exactly 3,955 profile mappings are rationally infeasible.
Transported coefficient-by-coefficient Farkas checks. · All 30,845 rooted-orbit/profile instances. · proved - At least 25,278 mappings satisfy the integer root-cell relaxation.
25,278 transported witnesses with 783,618 checked nonnegative entries. · The root-cell relaxation only; no five-block realization follows. · computed - Every rooted orbit retains at least eight explicitly witnessed integer-feasible profiles.
Independent per-orbit aggregation of transported witnesses. · All 2,145 certified rooted skeleton orbits. · computed
Ruled out in this epoch- Exclude any complete rooted skeleton orbit using only the current root-cell equalities and ten root-pair penalty inequalities.
All 2,145 rooted orbits and all 30,845 necessary profiles. · Every orbit retains at least eight independently checked integer witnesses. · artifacts/root-pair-excess-coupling-global-20260808/independent-check.json · Add a strictly stronger proved constraint involving actual outside subsets, multiple point links, or full five-block compatibility. - Treat rationally feasible/integer-UNSAT Z3 statuses as certified profile exclusions.
Nine signatures covering 1,612 mappings. · No independently replayable integer-UNSAT certificate was produced. · The checker records these as unresolved_transports=1612. · Supply independently checked cutting-plane, LRAT/DRAT, or exhaustive integer certificates.
Open leads- Two-link amalgamation for adjacent points.
It retains actual link incidence and shared-block compatibility absent from the exhausted root-cell relaxation. · Bound and test the first canonical adjacent-link pairs in the 15-cycle/root-31 case. · high · open - Proof-producing OPB leaf solving.
It remains the most direct exclusion route once a pinned RoundingSat/VeriPB/CakePB stack is available. · Replay the saved tiny calibration bounds before attempting a hash-bound rooted leaf. · high · open - Outside-subset realizability refinement.
The 31-cell witnesses do not identify subsets of the ten outside points. · Encode one surviving profile using exact outside point/pair degrees and triple coverage, with independent SAT decoding. · normal · open
Continuation checkpointObjective: Determine whether exact compatibility of two adjacent forced optimal links prunes structures that survive every root-cell profile.
First action: Run jq '.partitions[] | select(.partition==[15]) | {matrix,root:(.root_orbits[]|select(.representative_mask==31))}' artifacts/joint-orbit-census-20260808/manifest.json, then hash-bind the selected adjacent vertex pair in a new two-link protocol.
Stop condition: Redirect if the first bounded canonical pairs are all feasible, a checked link-orbit census cannot be obtained, or SAT decoding/proof replay disagrees.
Next moves- Design the bounded two-link amalgamation protocol for an adjacent pair in the partition-[15], root-mask-31 skeleton.
- Enumerate canonical optimal C(14,4,2) links with full pair-multiplicity matrices and a checked orbit census.
- Test exact shared-pair block compatibility; independently decode SAT pairs and replay a proof for each UNSAT link-orbit pair.
- Keep the global 279-signature table only as a sound prefilter for stronger encodings.
Citations
Tool disclosureGPT-5.6 Sol principal designed, audited, implemented, and interpreted the epoch. Pre-existing GPT-5.6 Terra delegate memos supplied advisory experiment-design and challenger leads only; their claims were independently recomputed and they were not counted as validation. Deterministic tools: Python 3.12.3, Z3 4.13.0, exact Fraction arithmetic, jq, sha256sum, and the project run_experiment wrapper. No proof assistant or external solver certificate checker was used for the nine integer-UNSAT solver statuses, which remain unresolved.; orchestration: gpt-5.6-sol principal with gpt-5.6-terra delegates.
- Duration
- 1171.4s
- Review state
- not a result claim
- Attempt ID
covering-c1553-20260808-233239-3a9c96
Human review ledgerNo human review recorded.