Back to Research Overview
RESEARCH NOTE • QPSI-NOTE-2026-02
PHYSICAL QPU RESEARCH NOTE • EXPERIMENT V1.1

Compiler-Enabled Quantum Query Advantage for Candidate-State Search on Physical Quantum Hardware

AUTHORS: Q-Psi Research Team
DATE: August 17, 2026
BACKEND: ibm_marrakesh (156Q)
ABSTRACT

We report an empirical demonstration of compiler-enabled quantum query advantage for software-repair candidate-state search on a 156-qubit superconducting quantum processor (ibm_marrakesh). Across 9 frozen instances drawn from 7 distinct programming language ecosystems at N=4, N=8, and N=16, physical quantum execution under a strictly verified black-box oracle model achieved lower effective verifier queries than the classical expected black-box search cost across 9/9 individual cases and 3/3 problem sizes. At N=16, the 95% Wilson confidence interval upper bound on effective quantum queries (7.360) remains strictly below the classical expected baseline of 8.5 queries. All protocol specifications, pre-execution fairness audits, raw execution bitstrings, and adversarial independent audits are published under open science principles.

1. Research Question

Can authentic search state spaces derived by a software-repair compiler participate in a bounded quantum query advantage experiment on physical quantum hardware without ground-truth leakage or side-channel rank bias?

While theoretical quantum search algorithms (such as Grover's algorithm and amplitude amplification) guarantee asymptotic query speedups on unstructured databases, real-world compilation pipelines often introduce heuristic ordering or structural correlations that could trivialize classical search if exposed. This work evaluates whether a rigorous black-box search interface preserves demonstrable quantum query advantage on physical NISQ hardware when operating on real-world compiler outputs.

2. Q-Psi State-Space Compiler

The Q-Psi compiler translates discrete multi-file software bug state spaces into structured candidate subspaces. In this experiment, candidate sets of size N ∈ {4, 8, 16} were extracted from 9 real software repair problems spanning Python, Go, Java, Rust, TypeScript, PHP, Ruby, and C/C++.

3. Why Black-Box Fairness Matters & Pre-QPU Correction

In our initial v1.0 design, candidate state indices correlated with compiler rank scores, creating potential side-channel information. An independent adversarial audit identified this issue before physical hardware execution.

Protocol v1.1 resolved this vulnerability by enforcing:

  • Deterministic answer-independent candidate permutations (seeded by candidate hashes).
  • Strictly opaque candidate IDs (0 to N-1) presented symmetrically to classical and quantum arms.
  • 100.0% mechanical truth-table verification between classical and quantum oracle evaluators.

4. Physical IBM Quantum Execution & Frozen Protocol

Execution was performed on IBM Quantum's 156-qubit Heron processor, ibm_marrakesh, under runtime Job ID da1c7rkdedkc73eqs5mg. Each of the 9 frozen circuits was executed with 2,048 physical shots (18,432 physical shots total).

5. Experimental Results

Problem Size (N)QubitsGrover kPhysical P_succQuantum Q_effClassical Q_classAdvantageVerdict
N = 42197.09%1.0302.5002.43xPASS
N = 83277.08%2.5954.5001.73xPASS
N = 164341.99%7.1448.5001.19xPASS

Note: Classical baseline Q_class = (N+1)/2 represents the exact expected number of black-box verifier queries for classical random sampling without replacement. Quantum effective queries Q_eff = k / P_succ.

6. Statistical Confidence Analysis

At N=16, physical measurement yielded 2,580 successful shots out of 6,144 total shots (P_succ = 41.99%). The two-sided 95% Wilson score confidence interval is [40.76%, 43.23%]. Inverting this interval yields a 95% confidence bound on effective quantum queries:

Q_eff(95% high) = 3 / 0.4076 = 7.360 < 8.500

Because 7.360 < 8.5, the frozen statistical decision gate is fully satisfied.

7. What This Demonstrates & What It Does NOT Demonstrate

WHAT THIS DEMONSTRATES
  • Compiler-enabled quantum query advantage on physical superconducting hardware.
  • Opaque black-box verifier contract integrity without ground truth leakage.
  • Advantage across 9/9 individual cases and 3/3 problem sizes up to N=16.
WHAT THIS DOES NOT DEMONSTRATE
  • No end-to-end wall-clock speedup over classical computers.
  • No general computational supremacy or commercial advantage.
  • No advantage over every structure-aware heuristic algorithm.

8. Evidence & Reproducibility

IBM RUNTIME JOB ID: da1c7rkdedkc73eqs5mg
RAW RESULTS SHA256: 8f65edbe0cea3ffdc16f3ff89b07beaf6ab111019a556c1cde56b274c0e18ff2
PROTOCOL FREEZE COMMIT: 1bf15b85f7fd206fb985f037845ee6f0d9188e7f
INDEPENDENT AUDIT COMMIT: e8efd49b28ceca86a281b59455b45c27775857e0

9. Citation

@techreport{qpsi2026compilergrover,
  title = {Compiler-Enabled Quantum Query Advantage for Candidate-State Search on Physical Quantum Hardware},
  author = {{Q-Psi Research Team}},
  institution = {Q-Psi Independent Quantum Research, The Oneness Project},
  year = {2026},
  month = {August},
  number = {QPSI-NOTE-2026-02},
  url = {https://qpsi.vercel.app/research/compiler-grover-query-advantage}
}