Workloads where QPC reports the same objective number as an external reference—published paper, classical baseline, or inequality bound—with auditable IBM (or other) job IDs.
D-Wave “Proof of Quantum Work” blockchain (Leap demo) is not on this list: it measures quantum-hash consensus on annealers, not folding energy, QUBO cost, or ICC. Different subject—results are not comparable rows.
Dedicated comparison program: same published D-Wave demo classes executed on IBM Heron via QPC — feature selection, bin packing, and fleet routing. Not a bake-off; a clarification: what you can do on quantum optimization, QPC can deliver on gate hardware too.
Special Report → Joint Structure Challenge →Published industry-style business tasks replayed on IBM Heron gate QPUs with QPC polycontextural architecture. Same metrics as external studies; auditable job IDs.
| Status | Task | Metric | QPC result | Report |
|---|---|---|---|---|
| ✓ Live | Ribeiro 2026 — Cerrado carbon portfolio (arXiv:2602.09047) | Portfolio score | Decoder 6.27 vs greedy 5.53 · ibm_fez ZNE | Overview · Technical |
| ✓ Live | D-Wave open demos — MIQUBO, binpack, MVRP (Cases 1–3) | QUBO energy gap vs reference | Gap 0.00 · ibm_fez Heron 152Q · jobs on record | Special Report |
| ● Next | D-Wave Mendeley 45 QUBO instances (MaxCut / TSP / Knapsack) | QUBO energy vs published NL-Hybrid | Planned — same Hamiltonian on IBM | Catalog A2 |
| ● Next | D-Wave Taillard flow-shop scheduling vignette | Optimality gap vs best known | Planned — supply-chain narrative | Catalog B1 |
| ● Next | LR-QAOA weighted MaxCut (cross-vendor IBM study) | Approximation ratio r | Planned — multi-context on Heron | Catalog A1 |
| # | External baseline | Metric | QPC | Report |
|---|---|---|---|---|
| 1 | IonQ / Kipu BF-DCQO (mastoparan I) | Energy after repair (↓ better) — not advantage claim | QPC −9.74 · random −10.65 · IonQ ref −8.70 · raw IonQ −4.19 | Protein |
| 2 | Controlled K=1 | Held-out fit | K=4 > K=1 (same resources) | QQ pilot |
| 3 | Classical KS bound | Witness vs bound | Heron Fez + mitigation | KS |
| 4 | Single-context baseline | ICC, FSP | Multi-context PFQM | PFQM |
| 5 | Classical per-context optimum (same grid) | QAOA objective J | proprietary (NDA) | GIC Case 3 JSON |
| 6 | IBM console audit | Job IDs, tiers T1–T4 | 4→24 contexts | Open instance |
| 7 | Separable control (same backend) | ICC gap (coupled − 3-job) | Pass toy 3/3; Cerrado; Heron boundary | Separability |
| 8 | E8 sphere packing (known 8D optimum — calibration only) | 8 parallel certificate checks; one IBM job | 152Q · job d8o47bjqv2lc7389jo60 · plain-language report |
E8 test |
● Separability falsification — ICC gap coupled vs independent IBM jobs; auditable on ibm_fez. Not an optimization bake-off.
Architecture program hub Joint Structure Challenge Separability results
● Transjunction ON vs OFF / ICC — completed on toy + Cerrado; see separability report.
● Cross-hardware JSC replication (IonQ / Quantinuum) — L5 open.
qaoa_best_objective_during_search, qaoa_objective_on_final_noiseless_histogram, classical baseline in output JSON.docs/QPC_COMPARABLE_BENCHMARKS_SHEET.mdSuggested standard add-ons: Qiskit MaxCut (approximation ratio), small Gset QUBO vs tabulated best—same metric as literature, 1-context vs N-context on IBM.