Real-world business tasks — on gate QPUs, with QPC

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.

Two kinds of quantum business computing

Annealers (e.g. D-Wave)

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.

Gate QPUs (IBM Heron — QPC)

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.

What we are proving (and what we are not)

We proveWe 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

How QPC computes (public overview)

QPC does not publish proprietary gate schedules or internal control secrets. The public architecture is:

[ Contexture A: objective 1 — cost + entanglement on block A (e.g. 52 qubits) ] ↓ transjunction (aligned business indices) [ Contexture B: objective 2 — cost + entanglement on block B ] ↓ transjunction [ Contexture C: objective 3 — cost + entanglement on block C ] ↓ measure portfolio register → classical decoder → portfolio score

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.

Flagship case: Brazilian Cerrado carbon credits

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.

The business job (plain language)

What “success” means here

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.

Auditable IBM Quantum jobs (QPC heron156 + ZNE)

Stepλ (noise fold)IBM job IDDecoder portfolio score
QPC polycontextural (156Q)1d8gj6t1e8nrc73bg1eh06.267
Gate folding2d8gj7042upec739jrgjg6.263
Gate folding3d8gj73dv8cos73f37hbg5.886
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.

Workflow contrast (why this matters vs typical vendor hybrids)

PatternSubmissionsCerrado 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

Relation to D-Wave’s public story

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.

Honesty & reproducibility

More comparable rows (same metric, external baseline)

We maintain a growing list of vendor-adjacent tasks with open data — comparison, not competition:

Cerrado portfolio ✓ VW Beijing traffic ✓ IonQ protein (IDWKK) Coupling ablation D-Wave Mendeley QUBOs (planned) Taillard flow-shop (planned)
Comparable benchmarks Vendor replication catalog QPC preprint (Zenodo) Highlights report