Answers are grounded in this paper and cite their sources.
Examines an abstention-aware audit technique for quantum execution outcome accuracy, indicating enhanced reliability.
Neuro-Core is an abstention-aware audit architecture for ordered-shot records returned by managed quantum-execution systems. Aggregate histograms preserve outcome frequencies but discard shot order. Neuro-Core addresses this evidential blind spot by combining canonical record admission, explicit provenance and sufficiency checks, an ordered-record statistical route, closed decision authority, calibrated abstention rules, and byte-exact replay bindings. The empirical evidence comprises two frozen hardware campaigns on ibm_kingston using a fixed 20-qubit GHZ circuit and layout. At the campaign-frozen serial-axis level of 0.05, configured zero repetition delay was serial-positive in 22 of 24 blocks, compared with 3 of 24 matched provider-default blocks, with a median paired effect ratio of 5.5 across three sessions. The frozen composite endpoint is nevertheless retained as FAIL because its preregistered specificity ceiling was exceeded by one block. At the later Bonferroni per-axis level of 0.025, the same records yield 21 of 24 against 1 of 24; this reanalysis does not amend the frozen campaign record. An order-shuffle control leaves every canonical histogram unchanged while reducing zero-delay detection from 22 of 24 to 2 of 24, demonstrating that the observed evidence is carried by shot order rather than by aggregate counts. A preregistered 112-block specificity campaign found no material departure from the intended null behaviour in the tested configuration. A separate specification-driven implementation reproduced all estimator values to 10^-16 and 47 of 48 classifications, with the single disagreement occurring at a Monte Carlo decision boundary. A development-mode replay of 288 archived units through the production engine exercises the abstention path directly: when an authorised serial-witness seed is absent, the engine does not reconstruct or substitute one and returns ABSTAIN through the corresponding admitted row of the closed decision table. The work does not certify computational correctness, establish a physical noise mechanism, benchmark general device capability, or claim cross-backend or cross-workload generalisation. Implementation conformance has not yet been externally tested, decision-grade standing is not claimed, and the marginal axis remains descriptive pending independent calibration and power assessment. This record contains the first public preprint of the work.
No takes yet. Share an insight, caveat, or question.
Emre Koçana (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: