Auditable Multi-Context Quantum Orchestration on IBM Heron
Protein Comparison, Controlled K-Ablations, and Open-Instance Scale Evidence
Abstract. Quantum Polycontextural Computing (QPC) is a hardware-agnostic orchestration layer that maps multiple labelled problem contexts onto one gate-model execution on commercial superconducting QPUs. We report three auditable hardware studies on IBM Heron (ibm_fez, ibm_marrakesh) under the IBM Open plan, each with published job identifiers and JSON artifacts: (i) a 46-qubit mastoparan-I folding pilot compared to a published trapped-ion BF-DCQO energy reference (−8.70); QPC Tier-D repair reaches −9.74 in one pass with eight recoverable job IDs; (ii) a controlled K-ablation on 16 qubits where polycontextural K=4 fits held-out Wang–Busemeyer joint statistics better than a matched K=1 control (bootstrap one-sided p<0.0005 on total-variation and KL); (iii) thirteen Runtime jobs demonstrating co-resident multi-context layouts from 4×39Q through 24 contexts in six cluster submissions. A supplementary transjunction ON vs OFF ablation at 3×26Q shows bridge-local inter-context correlators increase by |Δ|>0.05 when ring coupling is enabled, after QPC readout mitigation on bridge qubits only; full-width correlators do not separate conditions. We state limitations explicitly: QPC is not a new QPU; protein and cognition results are selective comparisons, not universal advantage claims; open-instance entropy alone does not signature the architecture.
Keywords: multi-context orchestration, IBM Quantum, auditable benchmarks, transjunction coupling, quantum chemistry pilot
1. Introduction
Enterprise and research workloads often require several scenarios at once—risk views, cost and carbon constraints, supply-chain hypotheses, or multiple experimental conditions. Mainstream gate-model stacks typically serialize these as separate circuits or flatten them into a single Boolean narrative per run. Quantum Polycontextural Computing (QPC) is an orchestration and logical-architecture layer that keeps multiple formal contexts co-resident on one chip width where the layout permits, coupled by transjunction gates when the workload requires cross-context structure.
This note is deliberately narrow: an 8–12 page hardware evidence summary with honest scope, not a complete theory monograph nor a claim of a new quantum processor. Every headline result below is tied to (a) a defined metric, (b) a stated baseline or control, and (c) verifiable IBM Quantum Runtime job IDs.
What we claim.
- QPC can plan, submit, and archive multi-context workloads on IBM Heron 156Q-class devices with reproducible JSON schemas and public job IDs.
- On three independent lines—published folding energy, controlled K-ablation on empirical joints, and bridge-local coupling metrics—we report comparisons where the metric and baseline are explicit.
What we do not claim.
- QPC does not replace IBM Quantum, Qiskit, or hardware vendors.
- High shot entropy or unique outcome counts from deep circuits are not unique signatures of QPC.
- We do not assert human-cognitive mechanisms, universal quantum advantage, or that no other architecture could match our pilots.
2. Methods
2.1 Context layouts and transjunctions
A context is a labelled register block on the device graph with its own morphogrammatic evolution. Transjunctions are two-qubit bridges linking aligned qubits across context boundaries. The orchestrator emits a single QASM submission per cluster job when contexts are co-resident.
2.2 Software stack
Runs use Qiskit Runtime SamplerV2 on IBM Open (open-instance). A QPC noise reducer is enabled where documented. Artifacts use versioned JSON schemas (qpc_benchmark_v1, qpc_coupling_ablation_v2, protein and QQ archives).
3. Experiment I — Mastoparan I vs published BF-DCQO
Sequence IDWKKLLDAAKQIL (mastoparan I) on 46 qubits on ibm_fez. Published reference Eref = −8.698 (IonQ/Kipu BF-DCQO, arXiv:2604.26861). Eight Runtime jobs merged in qpc_protein_ALL_REFINED_v2.json.
| Tier | Best energy | Role |
|---|---|---|
| A — raw | +13.5 | Direct pool |
| B — geometry filter | +13.5 | Constraints |
| C — light consensus | −0.32 | Partial repair |
| D — repair pool | −9.74 | Primary headline |
| E — refined consensus | +16.48 | Alternate path |
| Eref (published) | −8.70 | External ceiling |
Table 1. Mastoparan I energy tiers (IBM Fez).
Full report: QPC_PROTEIN_IDWKK_FEZ.html
4. Experiment II — QPC-QQ K-ablation
16 qubits, K=1 vs K=4, Wang–Busemeyer Clinton–Gore joints. Parameters from order-blind marginals only; joints held out for evaluation. 18 SamplerV2 jobs on ibm_fez, 4096 shots.
| K | TVAB | TVBA | TVmean |
|---|---|---|---|
| 1 (control) | 0.2815 | 0.3216 | 0.3015 |
| 2 | 0.2672 | 0.2491 | 0.2582 |
| 4 (polycontextural) | 0.2479 | 0.2541 | 0.2510 |
Table 2. Hardware K-ablation (May 2026). Bootstrap n=2000: mean TV(K=1)−TV(K=4)=+0.0505, 95% CI [+0.0359, +0.0655], one-sided p=0.
Report: QPC_QQ_PILOT_REPORT.html
5. Experiment III — Open-instance scale
Thirteen Runtime jobs (512 shots, May 2026): T01 4×39Q (d8atngdmdd1s73b8p4fg), T02 12×13Q, T03–T04 clusters on Marrakesh, T05 24 contexts in six jobs. Establishes orchestration reach and audit trails—not, alone, architectural discrimination.
Report: QPC_OPEN_INSTANCE_BENCHMARKS.html
6. Supplement — Coupling ON vs OFF
Matched 3×26Q = 78Q, 1024 shots. Bridge-local ICC with readout mitigation:
| Backend | Coupled job | Δ full-width ICC | Δ bridge ICC (mit.) |
|---|---|---|---|
| ibm_fez | d8fh39ralsvc7391oe30 | −0.0033 | +0.0640 |
| ibm_marrakesh | d8fi333alsvc7391pvag | −0.0048 | +0.0690 |
Report: QPC_COUPLING_ABLATION.html
7. Discussion and data availability
Job IDs and JSON schemas are the reproducibility contract. Overview: qpc_highlights_report.html. Scripts in site release: qpc_protein_idwkk_fez.py, scripts/run_qpc_coupling_ablation.py.
References
- Wang & Busemeyer, J. Math. Psychol. 50 (2006) — quantum cognition dataset.
- IonQ/Kipu BF-DCQO, arXiv:2604.26861 (2026) — mastoparan Eref.
- QPC public reports, quantumpolycontextural.ai (2026).
- IBM Quantum Heron, quantum.ibm.com.
To submit on arXiv: compile arxiv/qpc_hardware_note.tex → PDF, or print this page (File → Print → Save as PDF). See docs/QPC_ARXIV_SUBMISSION_CHECKLIST.md.