Protected Quantum Observable Readout for Hybrid Workflows: A Refresh Primitive on IBM Kingston Heron r3 Quantum Processor

Y⊗Z Hybrid Decision Workflow: Refresh-Cadenced Re-injection as a Protected Runtime Primitive on IBM Kingston Author: Amit Brahmbhatt Organisation: Quantum-Clarity LLC Date: April 21, 2026 IBM Job ID: d7k3tm8kj84c73cddorg (publicly verifiable) Backend: ibm_kingston (IBM Heron r3, 156 qubits, heavy-hex topology) Predecessor records: DOI 10.5281/zenodo.18498540 (February 5, 2026) DOI 10.5281/zenodo.19478241 (April 9, 2026) DOI 10.5281/zenodo.19501961 (April 10, 2026) Plain Language Summary What we found: We designed a hybrid quantum-classical decision workflow around the Y⊗Z orthogonal stabilizer resource established in our prior records and asked a concrete question: how long can a classical controller wait before the structured quantum signal becomes too degraded to act on? We prepared the Y⊗Z stabilizer state on IBM Kingston using the same four-qubit module [13, 14, 15, 19] validated in prior QuantaCore campaigns, inserted variable delays from 0 to 25 microseconds, and measured the primary stabilizer observable |YZII| under four conditions. We defined an actionability threshold of |YZII| > 0.5 — the signal level above which a classical controller can reliably distinguish the structured observable from noise and act on it. The result is clear and operationally meaningful. Passive delayed readout becomes non-actionable by 15 microseconds (|YZII| = 0.479, below threshold). Passive protected hold degrades similarly, also falling below threshold at 15 microseconds (|YZII| = 0.314). By contrast, refresh-cadenced re-injection of the protected state — resetting all qubits at the midpoint and re-preparing the full Y⊗Z stabilizer — remains actionable at 15 microseconds (|YZII| = 0.611, above threshold). At 25 microseconds, all delayed conditions fall below threshold. Why this matters: This experiment bridges the physical characterisation of prior records into an operational hybrid-workflow context. The question is no longer "how long does the protected state survive?" but "how long can a controller use it?" The answer, on current Kingston hardware, is: passive readout and passive hold lose actionability by 15 microseconds, while refresh-cadenced re-injection extends actionability to at least 15 microseconds. This identifies refresh as the operationally viable strategy for hybrid workflows requiring structured observable access beyond the passive coherence window. More broadly, this result points toward a new class of computational advantage available on NISQ-era hardware — not by running longer or deeper circuits, but by actively managing the structured quantum signal that fragile computations depend on. Current NISQ devices are noisy, short-lived, and error-prone, which makes most quantum algorithms impractical at scale. But the Y⊗Z protected plane offers a different path: rather than fighting decoherence globally, it identifies a specific geometric subspace where structured information survives longer, and pairs it with a refresh protocol that periodically re-injects that structure before it degrades below usefulness. This is not quantum error correction in the fault-tolerant sense — it makes no claim to preserving arbitrary logical content indefinitely. What it does offer is something more immediately practical: a protected observable primitive that keeps a fragile structured signal actionable long enough for a classical controller to use it. On hardware where unprotected signals routinely wash out before the control loop can respond, that is a meaningful and demonstrable advantage. It suggests that hybrid quantum-classical workflows on near-term devices can be made more reliable not by waiting for better hardware, but by choosing the right geometric subspace, the right observable, and the right refresh cadence for the task at hand. Abstract We report a hybrid decision workflow experiment using the Y⊗Z orthogonal stabilizer resource on IBM Kingston (ibm_kingston, Heron r3, 156 qubits, heavy-hex topology), module [13, 14, 15, 19]. The experiment uses the proven 12-gate Y⊗Z stabilizer preparation from prior QuantaCore records and sweeps five delay times (0, 5, 10, 15, 25 microseconds) under four conditions, measuring the primary stabilizer observable YZII as the structured signal available to a classical controller. Four conditions were tested: C1 (Immediate reference): Y⊗Z state prepared and measured without delay. Establishes full-strength baseline. Stable at |YZII| ≈ 0.68 across all delay points, confirming coherent observable mapping throughout the experiment. C2 (Passive delayed readout): Y⊗Z state prepared, delay inserted, measured. No protection protocol during delay. Decays from 0.665 at tau = 0 to 0.479 at tau = 15 microseconds (below threshold) and 0.342 at tau = 25 microseconds. C3 (Protected hold): Y⊗Z state prepared, delay inserted, measured. The Y⊗Z geometry provides passive coherence protection during the delay. Decays from 0.670 at tau = 0 to 0.314 at tau = 15 microseconds (below threshold) and 0.074 at tau = 25 microseconds. Underperforms passive delayed readout at late times, consistent with gate overhead interacting with decoherence. C4 (Protected refresh): Y⊗Z state prepared, delay tau/2 inserted, all qubits reset and full Y⊗Z state re-prepared, delay tau/2 inserted, measured. Decays from 0.680 at tau = 0 to 0.611 at tau = 15 microseconds (above threshold) and 0.374 at tau = 25 microseconds. Actionability threshold: |YZII| > 0.5 — defined as the signal level at which a classical controller can distinguish the structured observable from noise with confidence sufficient to gate a workflow action. Key result: At tau = 15 microseconds, C2 and C3 are both non-actionable (✗), while C4 alone remains actionable (✓). This is the first demonstration in the QuantaCore program of a concrete hybrid-workflow criterion — controller actionability — that distinguishes refresh-cadenced re-injection from both passive delayed readout and passive protected hold. Summary table: | tau (µs) | C1 |YZII| | C2 |YZII| | C3 |YZII| | C4 |YZII| | C2 | C3 | C4 | |---|---|---|---|---|---|---|---| | 0 | 0.678 | 0.665 | 0.670 | 0.680 | ✓ | ✓ | ✓ | | 5 | 0.674 | 0.565 | 0.616 | 0.612 | ✓ | ✓ | ✓ | | 10 | 0.698 | 0.544 | 0.560 | 0.631 | ✓ | ✓ | ✓ | | 15 | 0.681 | 0.479 | 0.314 | 0.611 | ✗ | ✗ | ✓ | | 25 | 0.671 | 0.342 | 0.074 | 0.374 | ✗ | ✗ | ✗ | The finding that passive protected hold (C3) underperforms passive delayed readout (C2) at late times is interpreted as evidence that gate overhead from the protection preparation interacts with decoherence at the 15–25 microsecond timescale in a way that accelerates rather than suppresses signal loss. This is consistent with observations in prior QuantaCore pilot experiments. Refresh-cadenced re-injection avoids this by re-injecting the full protected state at the midpoint delay, limiting each hold segment to approximately 7.5 microseconds — well within the high-fidelity window established in the April 10 hold-time record. This record establishes the fifth stage of a continuous QuantaCore research program: method disclosure (February 5, 2026), non-Markovian physical diagnosis (February 6, 2026), backend-adapted operational validation at scale (April 9, 2026), protected-plane lifetime characterisation (April 10, 2026), and hybrid-workflow actionability demonstration with refresh primitive (April 21, 2026, this record). 1. Experimental Background and Continuity This record is the fifth in a sequence of connected QuantaCore experimental disclosures. The first record (DOI 10.5281/zenodo.18498540, February 5, 2026) established the Y⊗Z parity-triangle consistency test and detected topology-dependent deviations from independence assumptions at 4.86σ significance on IBM Heron r2. The second record (February 6, 2026) extended that finding into a full non-Markovian reliability campaign with spatial correlations at 4.86σ, temporal memory at 2.8σ, and a characteristic environmental memory timescale of approximately 30 microseconds. The third record (DOI 10.5281/zenodo.19478241, April 9, 2026) validated the basis migration method at 116-qubit scale on IBM Kingston across 28 modules, with 89.39% average fidelity. The fourth record (DOI 10.5281/zenodo.19501961, April 10, 2026) directly measured the Y⊗Z protected-plane coherence lifetime as a function of hold time, establishing a high-fidelity window of approximately 15 microseconds, a non-Markovian revival near 25–30 microseconds, and a refresh advantage where reset-and-reprepare outperformed passive hold across much of the 2–50 microsecond window. The present record applies that physical characterisation directly to a hybrid-workflow context. Rather than asking "how long does the protected state survive?", we ask "how long can a classical controller act on it?" — introducing a concrete actionability criterion and demonstrating that refresh-cadenced re-injection extends the actionable window where passive strategies fail. 2. Experimental Design 2.1 Y⊗Z Stabilizer Preparation The 12-gate Y⊗Z stabilizer preparation sequence from prior QuantaCore records was used without modification. This sequence prepares a specific state inside the Y⊗Z orthogonal stabilizer subspace and is the same preparation used in Records 1–4. It is protected under U.S. Patent Application No. 19/643,807. Module [13, 14, 15, 19] on IBM Kingston was selected as the primary module, consistent with the highest-fidelity module identified in prior Kingston campaigns (F = 91.6% at tau = 0 in the April 10 record). 2.2 Four-Condition Design Condition Description Hypothesis C1 Immediate readout reference (no delay) Stable signal throughout; validates observable mapping C2 Passive delayed readout Decays; crosses threshold by ~15 µs C3 Protected passive hold Passive Y⊗Z geometry; expected to extend window C4 Protected refresh re-injection Reset + full re-preparation at tau/2; extends actionable window C3 and C4 use the same Y⊗Z preparation. The distinction between C2 and C3 is purely whether the Y⊗Z geometric structure provides passive coherence advantage during the idle period. C4 adds an active midpoint reset and full re-preparation. 2.3 Delay Sweep 0, 5, 10, 15, 25 microseconds — selected to span the high-fidelity window established in the April 10 record, with the critical crossover expected between 10 and 15 microseconds. 2.4 Actionability Threshold |YZII| > 0.5. Defined as the signal level above which a classical controller can reliably distinguish the structured observable from shot noise and act on it. With hardware standard deviation of approximately 0.035–0.045 per rep, this corresponds to a signal-to-noise ratio of approximately 11–14 at threshold. 2.5 Observable Primary Y⊗Z stabilizer observable YZII — the same observable used throughout Records 1–4. Physical qubits 14 (Y basis) and 13 (Z basis) on the 156-qubit Kingston register. 2.6 Execution Configuration Backend: ibm_kingston (IBM Heron r3, 156 qubits) Resilience level: 1 (TREX error mitigation) Twirling: Pauli gate and measure twirling (8 randomisations × 64 shots = 512 shots per circuit) Repetitions: 30 per condition per delay point Submission: Single batch job (600 PUBs total) Optimisation level: 1 3. Key Findings 3.1 Refresh Extends Actionability Where Passive Strategies Fail The central result is the tau = 15 microsecond crossover. At this delay: C2 (passive delayed readout): |YZII| = 0.479 — non-actionable (below threshold 0.5) C3 (protected passive hold): |YZII| = 0.314 — non-actionable C4 (protected refresh): |YZII| = 0.611 — actionable (margin +0.111 above threshold) This is the first QuantaCore result framed explicitly as a controller-actionability criterion, and it demonstrates that refresh-cadenced re-injection is the only tested strategy that preserves the hybrid controller's ability to act at the 15-microsecond delay. 3.2 C1 Stability Validates Experiment Integrity C1 (immediate readout reference, no delay) returns |YZII| in the range 0.671–0.698 across all five delay points. This stability confirms that the observable mapping, batch submission structure, and circuit execution are coherent throughout the experiment and that observed decays in C2–C4 are attributable to delay-induced decoherence rather than systematic calibration drift. 3.3 Passive Protected Hold Underperforms Passive Delayed Readout at Late Times C3 falls below C2 at tau = 15 microseconds (0.314 vs 0.479) and tau = 25 microseconds (0.074 vs 0.342). This is interpreted as evidence that the gate overhead of the Y⊗Z preparation circuit interacts with decoherence at the 15–25 microsecond timescale in a way that accelerates signal loss relative to the simpler passive delayed readout circuit. This finding is consistent with prior QuantaCore pilot experiments in which protected hold introduced systematic overhead that outweighed the geometric protection benefit at longer delays. It does not invalidate the protection claim — C4 demonstrates that the protective geometry is still useful when combined with refresh — but it establishes that passive hold alone is not sufficient at this timescale on current hardware. 3.4 Actionable Window All conditions are actionable from tau = 0 through tau = 10 microseconds. The actionable regime for refresh-cadenced re-injection (C4) extends to tau = 15 microseconds. At tau = 25 microseconds, all delayed conditions fall below threshold. The practical actionable window for the hybrid controller using C4 is therefore approximately 0–15 microseconds on IBM Kingston module [13, 14, 15, 19] under the current experimental configuration. 3.5 Connection to Prior Hold-Time Record The C4 midpoint delay of 7.5 microseconds (half of tau = 15 microseconds) is well inside the F > 80% high-fidelity window established in the April 10 record (~15 microseconds). This explains why C4 remains actionable at tau = 15 microseconds: each hold segment is short enough to remain in the high-fidelity regime, and the midpoint re-injection reinitialises the protected state before significant degradation accumulates. The refresh strategy is therefore not arbitrary — it is directly calibrated to the physical hold-time window demonstrated in the prior record. 4. IP Statement The Y⊗Z basis migration procedure and stabilizer preparation sequence implemented in this experiment are protected under U.S. Patent Application No. 19/643,807 (filed April 10, 2026), which claims benefit of Provisional Application No. 63/952,786 (filed January 2, 2026). This dataset constitutes an application-layer demonstration of the protected-plane resource in a hybrid quantum-classical decision workflow. Implementation scripts are proprietary and are not included in this dataset. 5. Reproducibility and Verification IBM Quantum job ID: d7k3tm8kj84c73cddorg — all results independently verifiable at https://quantum.ibm.com/jobs/d7k3tm8kj84c73cddorg Files in this record: yz_hybrid_decision_20260421_202158.json — complete per-condition, per-delay results including raw expectation values (30 reps each), mean, standard deviation, variance, actionability classification, and sign consistency statistics for all four conditions across all five delay points yz_hybrid_decision_summary.png (if included) — four-panel visualisation: |YZII| vs delay for all conditions, actionability threshold overlay, per-condition actionability classification table, and C1 stability plot 6. Acknowledgements IBM Quantum for hardware access and QPU credits on ibm_kingston. © 2026 Amit Brahmbhatt, Quantum-Clarity LLC. Data: CC BY 4.0.