Quantum Midi Posse Reports 96-Qubit Structured-Output Benchmark on IBM Hardware

Insider Brief

• Quantum Midi Posse announced results from a 96-active-qubit benchmark on IBM quantum hardware that the company says preserved structured output patterns under native-bridge NISQ execution without error correction or post-selection.
• The benchmark tested the firm’s proprietary Quantum State Command Encoding (QSCE) framework across six simultaneous 16-qubit regions using native cross-region bridge operations on IBM’s Marrakesh superconducting quantum system.
• The company said the experimental results suggest structured quantum output may remain distinguishable from multiple control configurations.

PRESS RELEASE — Quantum Midi Posse, an independent quantum research and intellectual property firm founded by Frank Angelo Drew, announces the completion of the Madmartigan Global Native-Bridge benchmark, a 96-active-qubit structured-output benchmark executed on IBM superconducting quantum hardware.

Quantum State Command Encoding (QSCE) is the architecture underlying the Madmartigan benchmark. QSCE treats quantum state preparation, phase structure, entanglement, and measurement as a command-bearing substrate. In the QSCE framing, state preparation functions as code, measurement collapse functions as execution, and quantum correlation functions as control. The Madmartigan benchmark tests whether this architecture can preserve reference-specific structured output under real NISQ hardware pressure.

The benchmark extends prior 96-active-qubit tiled Madmartigan work by testing a stricter condition: whether the original seed-1337 16-qubit Madmartigan structured-output band could survive a single global 156-qubit hardware execution with 96 active qubits, six simultaneous 16-qubit regions, and native cross-region bridge operations inserted between tile regions.

The final benchmark used the T6 rank-2 physical tile layout on IBM Marrakesh, preserved the exact original 16-qubit Madmartigan tile kernel, used 4096 shots per execution, and used no quantum error correction or post-selection. The native-bridge layer selected the available active-topology cross-region bridge edges, corresponding to physical qubit pairs (97,87) and (86,85). The raw exact-kernel native-bridge circuit transpiled to depth 654 with 1241 CZ gates, 2653 SX gates, 2496 RZ gates, 156 measurements, and 72 barriers.

Across five hardware executions, the raw bridged Madmartigan circuit produced repeatable Madmartigan-reference structured-output preservation. The aggregate raw result across 30 tile-level observations was:

• Mean F_XEB: 1.106965
• Mean HOG: 0.652515
• Mean Shannon entropy: 11.886623 bits per 16-qubit tile
• Mean IPR: 3661.616
• Positive F_XEB: 30/30 tile-level observations
• HOG above 0.55: 30/30 tile-level observations

The benchmark was then tested against a three-control ladder. The exact-scaled generic RCS native-bridge control matched the raw circuit in CZ count and barrier count, nearly matched depth, and preserved its own-reference structure, but collapsed near baseline against the Madmartigan reference with mean Mad-ref F_XEB = 0.053016 and mean Mad-ref HOG = 0.491268.

A phase-scrambled native-bridge control likewise preserved its own scrambled reference structure but failed to reproduce the Madmartigan band, with mean Mad-ref F_XEB = -0.015775 and mean Mad-ref HOG = 0.495589.

A partial-entanglement native-bridge ablation produced a separate high-attractor QSCE-derived own-reference mode, with mean own-reference F_XEB = 3.202684 and mean own-reference HOG = 0.762719, while retaining only weaker Madmartigan-reference overlap with mean Mad-ref F_XEB = 0.150540 and mean Mad-ref HOG = 0.549089.

Statistical analysis over tile-level observations showed strong raw-versus-control separation. Raw bridged Madmartigan versus exact-scaled generic RCS produced Cohen’s d = 2.399974 for F_XEB and d = 4.015554 for HOG, with Welch p-values of 1.753022 x 10^-10 and 7.374341 x 10^-18, respectively. Raw versus phase-scrambled produced Cohen’s d = 2.596101 for F_XEB and d = 4.106335 for HOG. Raw versus partial-entanglement ablation produced Cohen’s d = 2.179171 for F_XEB and d = 2.433099 for HOG. Bootstrap confidence intervals for all primary raw-minus-control F_XEB/HOG differences excluded zero.

The broader implication of the Madmartigan Global Native-Bridge benchmark is that NISQ quantum utility may not be limited to probabilistic sampling or waiting for large-scale fault-tolerant logical qubits. In the QSCE framework, the measured quantum distribution is treated as a structured signal surface: state preparation functions as code, collapse functions as execution, and correlation functions as control.

This benchmark also sits within a larger QSCE evidence chain. In prior IPCM work, QSCE output was used as a quantum-to-classical signaling pathway, where a measured dominant quantum state was decoded and routed into a classical network message. Madmartigan addresses the upstream benchmark problem: whether structured, reference-specific output can survive larger physical NISQ execution pressure. IPCM addresses the downstream systems problem: whether quantum output can be decoded into a classical signaling or command pathway.

Taken together, the Madmartigan and IPCM results point toward a possible alternate path for quantum utility: deterministic or near-deterministic signaling through engineered structured-output bands, rather than treating quantum hardware output only as probabilistic samples requiring post-hoc interpretation. The present benchmark strengthens the physical-NISQ substrate behind QSCE’s broader quantum-to-classical signaling architecture.

“This benchmark points to a different lane for quantum utility,” said Frank Angelo Drew, founder of Quantum Midi Posse. “The field has largely treated NISQ output as probabilistic sampling while waiting for fault-tolerant logical qubits. Madmartigan shows that engineered quantum structure can survive directly on physical hardware as a reference-specific output band. When viewed alongside our IPCM quantum-to-classical signaling prototype, the implication is clear: QSCE is developing toward deterministic or near-deterministic signaling through structured collapse surfaces, not merely better sampling.”

The work has also been shared for external SME technical review with Dr. Joseph Mitola III, an IEEE Life Fellow widely known for foundational work in software-defined radio and cognitive radio systems. Dr. Mitola has previously reviewed the broader QSCE architecture and its potential defense and quantum-systems implications, and the finalized Madmartigan benchmark papers have been prepared to support deeper technical review.

The benchmark paper and supporting documentation are being made available through Zenodo as part of Quantum Midi Posse’s public technical record. Statistical appendix files, control-separation tables, run-level summaries, topology/profile summaries, and supporting CSV outputs are available upon request. Executable circuit artifacts, QPY/QASM3 files, and source-level reproduction materials may be made available through protected review channels to preserve intellectual property integrity.