Quantum Polycontextural Computing (QPC) takes published industry problems — the same multi-objective optimization stories vendors talk about — and runs them on IBM Heron gate hardware using a unique polycontextural architecture. This is comparison and demonstration, not a claim to “beat” D-Wave, IonQ, or anyone else on their hardware.
Problems are cast as QUBO / BQM. A quantum annealer searches low-energy states of that single objective landscape. Hybrid tools often loop classical optimizers around repeated annealer calls.
Typical marketing: logistics, inventory, scheduling — often supported by surveys and vignettes, not always by open job IDs on your exact instance.
Problems are expressed as circuits (QAOA-style cost layers, mixers, measurements). QPC adds several objective “contextures” on one device layout, coupled by transjunctions, in one submission where possible.
We report the same business metric as the published study (portfolio score, QUBO cost, energy, etc.) plus auditable IBM job IDs.
Your D-Wave business PDF (Wakefield survey, 2025) lists problem themes (supply chain, routing, ROI). It does not ship matrices, scores, or job traces — so it cannot be replayed. QPC uses open academic and vendor datasets instead.
| We prove | We do not claim |
|---|---|
| Real business structure (multi-objective, spatial synergies, cardinality) can run on gate hardware as one polycontextural workflow | QPC is faster or “better” than a D-Wave Advantage annealer on the same QUBO |
| Same numeric metric as a published external reference, with IBM job IDs | Headline scores from other vendors’ papers transfer line-for-line without matching their full pipeline (width, ZNE, instance size) |
| Classical–quantum hybrid decoding (noise reducer + portfolio repair) is industry-standard assembly around quantum samples | Every measured bitstring must already satisfy hard constraints without classical help |
QPC does not publish proprietary gate schedules or internal control secrets. The public architecture is:
On IBM Heron (156 qubits), our Cerrado deployment uses 3 × 52 qubits = 156Q — using the full chip width for balanced contextures, not forcing an 88-qubit copy of someone else’s paper width.
External reference: Ribeiro 2026, arXiv:2602.09047 — QAOA + zero-noise extrapolation on IBM for municipality portfolio selection in Goiás (Brazilian Cerrado). Open data: github.com/hgribeirogeo/qaoa-carbon-cerrado.
evaluate() function (we verify classical greedy against their reported baseline on the large instance).Verified on ibm_fez (June 2026): QPC polycontextural run + industry-standard decoder beats classical greedy on the same metric.
Classical greedy: 5.528 · QPC (decoder + ZNE extrapolation): 6.272 · Improvement: +13% on this instance.
Reference paper (different width, their ZNE pipeline): greedy 44.42, QAOA+ZNE mean ~58.47 at n=88 — cited for context, not as a direct race on identical shots.
| Step | λ (noise fold) | IBM job ID | Decoder portfolio score |
|---|---|---|---|
| QPC polycontextural (156Q) | 1 | d8gj6t1e8nrc73bg1eh0 | 6.267 |
| Gate folding | 2 | d8gj7042upec739jrgjg | 6.263 |
| Gate folding | 3 | d8gj73dv8cos73f37hbg | 5.886 |
| ZNE extrapolated (λ→0) | — | — | 6.272 |
Earlier single-run decoder success (no ZNE): job d8ga43tv8cos73f2rikg, score 5.908. Full tables, Aer dry-runs, and architecture baselines: Cerrado comparison report.
| Pattern | Submissions | Cerrado example |
|---|---|---|
| Sequential objective jobs (Classiq / multi-cloud style) | 3+ | Carbon QAOA, biodiversity QAOA, social QAOA → classical merge |
| Single weighted QUBO + QAOA (+ ZNE in paper) | 1 (+ folds) | Ribeiro baseline; we reach ~6.0 on 28Q Fez without 156Q layout |
| QPC polycontextural | 1 (+ optional ZNE folds) | Three contextures + transjunctions on ibm_fez 156Q → decoder score 6.27 |
D-Wave’s survey PDF and website describe the same class of pain: hard combinatorial decisions under competing KPIs (cost, ESG, risk, geography). That narrative is real; the gap is reproducible numbers.
One-sentence positioning: QPC demonstrates that published real-world optimization structure — of the kind quantum vendors market to enterprises — can be executed as one polycontextural IBM submission with measurable portfolio improvement over classical greedy, using open data and verifiable job IDs.
results/cerrado_compare_heron156_fez_v3_zne.json, results/cerrado_compare_heron156_fez_v3_decoder.json.We maintain a growing list of vendor-adjacent tasks with open data — comparison, not competition: