{ "audit_task_id": "QPSI_MASTER_EVIDENCE_AUDIT_AND_FREEZE_2026_08_18", "freeze_date": "2026-08-18", "version": "2.0.0-audited-freeze", "program_totals": { "total_physical_shots": 601704, "completed_audited_experiment_families": 15, "production_qpu_jobs": 19, "ibm_runtime_quantum_seconds": 288.38, "primary_backend": "ibm_marrakesh", "processor": "IBM Heron r2 (156 programmable qubits)" }, "claims": [ { "claim_id": "CLAIM_COMPILER_GROVER_QUERY_ADVANTAGE", "short_claim": "Bounded physical quantum query advantage for compiler-synthesized search problems.", "status": "SUPPORTED", "publication_tier": "TIER_A", "experiment_ids": [ "QPSI_COMPILER_GROVER_V1_1" ], "evidence_hashes": { "raw_results_sha256": "8f65edbe0cea3ffdc16f3ff89b07beaf6ab111019a556c1cde56b274c0e18ff2" }, "allowed_wording": [ "Q-Psi demonstrated compiler-enabled physical quantum query advantage on structured search problems on IBM Heron hardware under a black-box verifier model.", "Quantum query reduction factors of 2.43x (N=4), 1.73x (N=8), and 1.19x (N=16) observed on physical quantum hardware (ibm_marrakesh)." ], "prohibited_wording": [ "General quantum speedup for arbitrary software repair.", "End-to-end wall-clock compiler runtime advantage.", "General quantum advantage." ], "qualification": "Valid only under the query complexity metric on tested structured oracle families; does not establish wall-clock speedup or general quantum advantage.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_DYNAMIC_BV_QUERY_ADVANTAGE", "short_claim": "Bounded physical quantum query complexity advantage on Dynamic Bernstein-Vazirani scaling series.", "status": "SUPPORTED", "publication_tier": "TIER_A", "experiment_ids": [ "QPSI_DYNAMIC_BV_QUERY_ADVANTAGE" ], "evidence_hashes": { "evidence_sha256": "72e8773865cdcbba302f24ae2bfc136c3a0cf4878a2493db0828a104508fa1ba" }, "allowed_wording": [ "Demonstrated statistically significant physical quantum query scaling advantage on Dynamic Bernstein-Vazirani problem across N=4..16 qubits on IBM Marrakesh.", "Quantum query scaling slope (0.153) is statistically superior to classical exhaustive bound (0.696, t = -30.65, p < 1e-6)." ], "prohibited_wording": [ "General quantum supremacy.", "General quantum advantage for arbitrary computational problems." ], "qualification": "Restricted to the tested oracle complexity model on synthetic scaling series.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_MCM_QUBIT_REUSE_WIDTH_REDUCTION", "short_claim": "Mid-circuit measurement and reset reduces physical qubit footprint by median 78.1% with non-inferior observable quality.", "status": "SUPPORTED_WITH_QUALIFICATION", "publication_tier": "TIER_A", "experiment_ids": [ "QPSI_MCM_QUBIT_REUSE_V1" ], "evidence_hashes": { "raw_results_sha256": "ec0f2cea637f433f9d134d05a7168b3114b4526c2f5483288efbdea01d83e8e4" }, "allowed_wording": [ "Dynamic mid-circuit qubit reuse compiler passes reduce active physical register width by up to 87.5% (median 78.1%) on streaming and bounded-treewidth circuits without observable degradation (6/6 non-inferior cases).", "Evaluated 16 logical-qubit workloads on 2 to 3 physical slots on IBM Marrakesh." ], "prohibited_wording": [ "16-qubit quantum computation generally requires only 2 qubits for all algorithms.", "Universal width reduction for arbitrary quantum circuits." ], "qualification": "Applicable strictly to circuits with sequential liveness graphs and streaming communication topologies; arbitrary non-local entanglement requires full register width.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_DATA_FOUNDRY_V2_120QSEC_FEASIBILITY", "short_claim": "Physical-QPU-assisted code data generation pipeline feasibility demonstrated with 127.58 quantum seconds on Heron r2.", "status": "SUPPORTED_WITH_QUALIFICATION", "publication_tier": "TIER_A", "experiment_ids": [ "QPSI_QUANTUM_DATA_FOUNDRY_CODING_V2_120QSEC" ], "evidence_hashes": { "raw_sample_sha256": "dfd11daa2f1f51b29e15019741363a3fd03672f8ccd562c52c4a762d656dc89f" }, "allowed_wording": [ "Physical QPU sampling on IBM Marrakesh (127.58s runtime, 332,768 shots across 4 jobs) causally selected candidate mutation states, yielding 93 verified positive code repair records (1.59 MB) passing rigorous classical unit and integration verifiers.", "Demonstrates mechanical end-to-end pipeline feasibility for quantum-assisted synthetic data generation." ], "prohibited_wording": [ "Quantum-trained LLMs achieve superior coding performance.", "Proven language model improvement from quantum synthetic data.", "Quantum data foundry generates 1000s of verified records per second without qualification." ], "qualification": "Feasibility demonstration only. The current throughput bottleneck is classical verification (158 candidates evaluated of 65k QPU states). Downstream LLM fine-tuning and model improvement were NOT tested.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_MQT_CROSS_COMPILER_BENCHMARK", "short_claim": "Independent cross-compiler benchmark on MQT suite shows mixed competitiveness against Qiskit O3 and TKET.", "status": "MIXED_OR_INCONCLUSIVE", "publication_tier": "TIER_A", "experiment_ids": [ "QPSI_MQT_CROSS_COMPILER_BENCHMARK_V1" ], "evidence_hashes": { "raw_results_sha256": "9d347393dd9f2b3c168ab706d2c4d55e0ff4ab1c65ea5091fb8b62527fc7e3e5" }, "allowed_wording": [ "On standard MQT benchmarks on IBM Marrakesh, Q-Psi achieves 18/18 valid compilations, ties or competes with Qiskit/TKET across most workloads, exhibits a large advantage over TKET on vqe_su2_n8, but remains physically disfavored vs Qiskit on qft_n6.", "Commercial benchmark classification: MIXED_COMPETITIVENESS." ], "prohibited_wording": [ "Q-Psi universally outperforms Qiskit or TKET.", "General compiler superiority across all quantum algorithms." ], "qualification": "Aggregate physical advantage over TKET (+0.0398 delta) is driven by TKET's mapping failure on vqe_su2_n8 (+0.2696); omitting vqe_su2_n8 results in -0.0062 delta. Q-Psi is physically disfavored vs Qiskit (-0.0067 delta).", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_DYNAMIC_ROUTING_CROSSOVER", "short_claim": "Dynamic ancilla routing improves gate depth but shows no physical fidelity crossover over unitary SWAP chains on Heron r2.", "status": "SUPPORTED_WITH_QUALIFICATION", "publication_tier": "TIER_B", "experiment_ids": [ "QPSI_DYNAMIC_ROUTING_CROSSOVER_V1" ], "evidence_hashes": { "raw_results_sha256": "6c8527a9854651585d5262b3c2399e612573906281ec030eef3be0005dbf83f0" }, "allowed_wording": [ "Dynamic routing structurally compresses 2Q depth from 73 to 2 on long distances (13 hops), but hardware reset and measurement noise prevent a physical Bell fidelity crossover on IBM Marrakesh, confirming pre-QPU compiler predictions." ], "prohibited_wording": [ "Dynamic routing always outperforms unitary routing on current hardware.", "Quantum routing advantage." ], "qualification": "Unitary routing physically wins on short and medium distances; long distance (13 hops) is statistically inconclusive (-0.0314 delta).", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_COMPILER_STAGE6F_OPTIMIZATION", "short_claim": "Hardware-fit compiler optimization achieves exact classical parity for small instances (N<=10) on Heron r2.", "status": "SUPPORTED_WITH_QUALIFICATION", "publication_tier": "TIER_B", "experiment_ids": [ "QPSI_STAGE6F_PHYSICAL_OPTIMIZATION" ], "evidence_hashes": { "raw_results_sha256": "3f3b4c3982a267aa7fbf297adc4171ded2c8b1e24226a379df94bcfc53ac2783" }, "allowed_wording": [ "QAOA-based compiler optimization on IBM Marrakesh achieves 4/4 exact optimum hits on small compiler graphs (N<=10), matching classical branch-and-bound, but degrades on N=18..25 due to hardware noise." ], "prohibited_wording": [ "Quantum compiler optimization outpaces classical solvers.", "Quantum advantage for compiler optimization." ], "qualification": "Small graph parity only; large graphs degrade under unmitigated NISQ error.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_QUANTUM_DATA_FOUNDRY_CODING_V1", "short_claim": "Historical Data Foundry V1 proof-of-concept baseline.", "status": "SUPPORTED_WITH_QUALIFICATION", "publication_tier": "TIER_B", "experiment_ids": [ "QPSI_QUANTUM_DATA_FOUNDRY_CODING_V1" ], "evidence_hashes": { "raw_results_sha256": "9f7ac0139566025b7b932df60f8fff1313698170b17ab9ffcd140a5e3ae1b998" }, "allowed_wording": [ "Initial feasibility demonstration of physical QPU selector on M=1024 space yielding 14 verified unique code repair records." ], "prohibited_wording": [ "Headline data generation result." ], "qualification": "Proof-of-concept run; superseded by V2 120-Qsec campaign.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_CALIBRATION_AWARE_LAYOUT_VS_QISKIT", "short_claim": "Q-Psi calibration-aware layout does not outperform strong Qiskit O3 best-of-8 seed baseline on Heron r2.", "status": "NOT_SUPPORTED", "publication_tier": "TIER_C", "experiment_ids": [ "QPSI_CALIBRATION_AWARE_LAYOUT_V1" ], "evidence_hashes": { "raw_results_sha256": "5c5b9dff337488cb35d4fcc2de0efcbe2e63c05f1264ef05c8bb006e17df0eb6" }, "allowed_wording": [ "Against a strong baseline of Qiskit O3 best-of-8 seeds, Q-Psi calibration-aware layout achieves 1 win, 2 losses, 3 inconclusive (mean delta -0.0106), demonstrating that Qiskit multi-seed search remains superior or competitive on Heron r2." ], "prohibited_wording": [ "Q-Psi calibration layout outperforms Qiskit layout on IBM Heron.", "Compiler layout advantage over Qiskit." ], "qualification": "Negative comparator study preserved for scientific integrity; demonstrates rigorous external baseline comparison.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_FRACTIONAL_GATE_LOWERING", "short_claim": "Fractional pulse gate lowering reduces pulse duration structurally but yields mixed physical fidelity improvements.", "status": "MIXED_OR_INCONCLUSIVE", "publication_tier": "TIER_C", "experiment_ids": [ "QPSI_FRACTIONAL_GATE_LOWERING_V1" ], "evidence_hashes": { "raw_results_sha256": "8a5cb81cf1b069baed93926ab75fcd9dd37a589c6d220aa41455293f2ebc23ae" }, "allowed_wording": [ "Fractional gate lowering yields structural gate duration reductions and 2/6 physical wins (N4_B, N8_B), with 4/6 cases inconclusive on IBM Marrakesh, supporting mixed physical signal." ], "prohibited_wording": [ "Universal fidelity improvement from fractional pulse lowering.", "Pulse-level quantum advantage." ], "qualification": "Only 2/6 cases were conclusive; physical advantage is workload-dependent.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_DIRECTED_EXECUTION_SELECTION", "short_claim": "Selective Pauli twirling reduces randomized compilation overhead by median 58.9% with mixed fidelity retention.", "status": "MIXED_OR_INCONCLUSIVE", "publication_tier": "TIER_C", "experiment_ids": [ "QPSI_DIRECTED_EXECUTION_SELECTION_V1" ], "evidence_hashes": { "raw_results_sha256": "2ae1632ebe9645be032fb2a29f60eb669ecc8bd8712d68a5837039b5bd79a97a" }, "allowed_wording": [ "Risk-mass layer selection selects only 41.1% of CZ layers for twirling on IBM Directed Execution beta mode; yields 1/4 win vs unmitigated RAW baseline and statistically inconclusive comparison vs full twirling." ], "prohibited_wording": [ "Selective twirling proven superior to full randomized compiling.", "Directed execution advantage." ], "qualification": "Selective vs full non-inferiority was not statistically established across 4 cases; retains mixed signal.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_RESTRICTED_SIMON_ALGORITHM", "short_claim": "Restricted Simon physical study achieves polynomial query reduction on recovered instances but inconclusive asymptotic advantage.", "status": "MIXED_OR_INCONCLUSIVE", "publication_tier": "TIER_C", "experiment_ids": [ "QPSI_RESTRICTED_SIMON_PHYSICAL_STUDY" ], "evidence_hashes": { "evidence_sha256": "aa22b8b023d03f9f64ccfc35b268f9dc3536fac149843938e67f0e381e4ef79a" }, "allowed_wording": [ "Physical execution on IBM Marrakesh recovered secrets on 6/12 instances (N=8..28) with polynomial query count reduction relative to classical bound, but noise-induced degradation on larger N prevents establishing asymptotic quantum speedup." ], "prohibited_wording": [ "Demonstrated asymptotic quantum speedup for Simon's algorithm.", "Unconditional physical quantum advantage." ], "qualification": "Inconclusive due to 6/12 recovery failures at higher N under hardware noise.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_STATIC_BV_MANTRA_EXPLORATORY", "short_claim": "Early exploratory benchmark on static BV and mantra state encoding.", "status": "EXPLORATORY", "publication_tier": "TIER_C", "experiment_ids": [ "QPSI_STATIC_BV_MANTRA_EXPLORATORY" ], "evidence_hashes": { "evidence_sha256": "6d24837313ca01b52f1d602ef0249f0bcc1b6714bb5ba974ec967153674d838f" }, "allowed_wording": [ "Exploratory hardware calibration run on static BV and mantra state encoding on IBM Marrakesh." ], "prohibited_wording": [ "Algorithmic quantum advantage on static BV." ], "qualification": "Exploratory early benchmark only.", "audit_status": "AUDITED_VERIFIED" }, { "claim_id": "CLAIM_MULTI_BACKEND_SELECTION", "short_claim": "Multi-backend retargeting across 3 IBM Heron r2 backends with 5/6 top-tier selection accuracy.", "status": "SUPPORTED_WITH_QUALIFICATION", "publication_tier": "TIER_B", "experiment_ids": [ "QPSI_MULTI_BACKEND_SELECTION_V1" ], "evidence_hashes": { "raw_results_sha256": "4218490b1044e1e99ddda8ffd278ab2f7525d400e7025c181f4810496f3ac576" }, "allowed_wording": [ "Across six canonical workloads and three IBM Heron r2 backends, Q-Psi successfully retargeted all 18 workload-backend combinations.", "Its frozen pre-QPU selection heuristic selected a backend in the physically best statistical tier for 5/6 workloads." ], "prohibited_wording": [ "Multi-provider portability.", "Cross-vendor compiler independence.", "General backend-selection superiority.", "Universal compiler advantage.", "General quantum advantage.", "Best backend selected on 5/6 workloads." ], "qualification": "Tested on three IBM Heron r2 backends from one provider and one processor family. This does not establish multi-provider portability, universal backend-selection superiority, general compiler advantage, or general quantum advantage.", "audit_status": "POST_FREEZE_ADDENDUM_SUPPORTED", "qpu": "ibm_fez, ibm_kingston, ibm_marrakesh (IBM Heron r2, 156 programmable qubits)", "job_ids": [ "da1uqsaein7c73bdtkm0", "da1uqtm3kjvs738783qg", "da1uqumg52gs73cm4v50" ], "shots": 36864, "quantum_seconds": 29.0 }, { "claim_id": "CLAIM_GENERAL_QUANTUM_ADVANTAGE", "short_claim": "General quantum advantage for arbitrary classical software or compilation tasks.", "status": "NOT_SUPPORTED", "publication_tier": "DISCLAIMED", "experiment_ids": [], "evidence_hashes": {}, "allowed_wording": [ "General quantum advantage is NOT established by Q-Psi research." ], "prohibited_wording": [ "Q-Psi has achieved general quantum advantage.", "Quantum computers outperform classical computers on general software tasks." ], "qualification": "Explicitly rejected across the entire research program.", "audit_status": "EXPLICITLY_DISCLAIMED" }, { "claim_id": "CLAIM_GENERAL_COMPILER_SUPERIORITY", "short_claim": "General compiler runtime or optimization superiority over industrial compiler frameworks.", "status": "NOT_SUPPORTED", "publication_tier": "DISCLAIMED", "experiment_ids": [], "evidence_hashes": {}, "allowed_wording": [ "General compiler superiority over industrial toolchains (Qiskit, TKET) is NOT established." ], "prohibited_wording": [ "Q-Psi is universally superior to Qiskit and TKET." ], "qualification": "Explicitly disclaimed; results show specialized advantages and strong baseline parity.", "audit_status": "EXPLICITLY_DISCLAIMED" } ], "historical_master_freeze_totals": { "freeze_date": "2026-08-18", "freeze_commit": "be74ad05187e148e2fc52309f4d7f57be3784157", "completed_audited_experiment_families": 14, "production_qpu_jobs": 16, "total_physical_shots": 564840, "ibm_runtime_quantum_seconds": 259.38 } }