We analyze 10.4 million calibration measurements from three IBM Heron superconducting quantumprocessors and identify two physically distinct classes of correlated multi-qubit degradationevents using only routine cloud-accessible telemetry. A column-wise permutation test reveals statisticallysignificant excess of simultaneous T1 crashes (42.8× expected at ≥ 10 simultaneous,p < 0.0001), simultaneous T2 crashes (4.8× at ≥ 10, p < 0.001), and simultaneous CZ gate-errorspikes (23.4× at ≥ 20, p < 0.0001; excess reaching 15,000× at ≥ 30). All three signals replicateacross backends (Fisher combined p < 10−7). Classifying high-crash snapshots by which parametersdegrade reveals two event types: Type A events degrade T1, T2, and CZ gate fidelity simultaneously(Cohen’s d = −0.94 for crashed-qubit T2), consistent with quasiparticle poisoning from high-energyparticle impacts; Type B events degrade T2 alone without affecting T1 or gate errors, consistentwith correlated dephasing noise. Type B events recover more slowly than Type A, persisting overmultiple calibration cycles with the same qubits remaining affected. T1-only events are rare acrossbackends, consistent with the expectation that relaxation degradation typically accompanies dephasing.Different qubits crash each time (mean Jaccard similarity ≤ 0.04), and readout fidelityremains unaffected in both event types. These correlated events—previously detectable only withdedicated particle detectors—are identifiable, classifiable, and operationally actionable from existingcloud calibration data without hardware modification.
We analyze 10.4 million calibration measurements from three IBM Heron superconducting quantumprocessors and identify two physically distinct classes of correlated multi-qubit degradationevents using only routine cloud-accessible telemetry. A column-wise permutation test reveals statisticallysignificant excess of simultaneous T1 crashes (42.8× expected at ≥ 10 simultaneous,p < 0.0001), simultaneous T2 crashes (4.8× at ≥ 10, p < 0.001), and simultaneous CZ gate-errorspikes (23.4× at ≥ 20, p < 0.0001; excess reaching 15,000× at ≥ 30). All three signals replicateacross backends (Fisher combined p < 10−7). Classifying high-crash snapshots by which parametersdegrade reveals two event types: Type A events degrade T1, T2, and CZ gate fidelity simultaneously(Cohen’s d = −0.94 for crashed-qubit T2), consistent with quasiparticle poisoning from high-energyparticle impacts; Type B events degrade T2 alone without affecting T1 or gate errors, consistentwith correlated dephasing noise. Type B events recover more slowly than Type A, persisting overmultiple calibration cycles with the same qubits remaining affected. T1-only events are rare acrossbackends, consistent with the expectation that relaxation degradation typically accompanies dephasing.Different qubits crash each time (mean Jaccard similarity ≤ 0.04), and readout fidelityremains unaffected in both event types. These correlated events—previously detectable only withdedicated particle detectors—are identifiable, classifiable, and operationally actionable from existingcloud calibration data without hardware modification.
Abstract Current quantum processor monitoring relies on individual qubit metrics (T1, T2, gate fidelity, readout error), missing degradation encoded in qubit-qubit relationships. Here we apply Weighted Jaccard (WJ) similarity analysis to track correlation architecture reorganization across three IBM quantum processors (ibm_fez, ibm_torino, ibm_marrakesh; 133–156 qubits each) over 30 days of calibration data. Full pairwise qubit correlation matrices are computed at each calibration snapshot and compared between temporal windows. All three backends show statistically significant architecture reorganization (WJ = 0.968–0.974, p < 0.001, permutation test, FDR-corrected; 95% bootstrap CI width < 0.012). The qubits participating most in correlation architecture reorganization show no association with conventional degradation indicators (rho = -0.051, p = 0.530), indicating that WJ detects a class of hardware change invisible to component-level monitoring. Cross-backend comparison reveals that architectural divergence between processors (~ 8–9%) exceeds within-processor temporal drift (~ 3%), suggesting hardware-specific correlation signatures.
Cloud quantum platforms give users access to many backends with different qubit technologies, coupling layouts, and noise levels. The execution of a circuit, however, depends on internal allocation and routing policies that are not observable to the user. A provider may redirect jobs to more error-prone regions to conserve resources, balance load or for other opaque reasons, causing degradation in fidelity while still presenting stale or averaged calibration data. This lack of transparency creates a security gap: users cannot verify whether their circuits were executed on the hardware for which they were charged. Forensic methods that infer backend behavior from user-visible artifacts are therefore becoming essential. In this work, we introduce a Graph Neural Network (GNN)-based forensic framework that predicts per-qubit and per-qubit link error rates of an unseen backend using only topology information and aggregated features extracted from transpiled circuits. We construct a dataset from several IBM 27-qubit devices, merge static calibration features with dynamic transpilation features and train separate GNN regressors for one- and two-qubit errors. At inference time, the model operates without access to calibration data from the target backend and reconstructs a complete error map from the features available to the user. Our results on the target backend show accurate recovery of backend error rate, with an average mismatch of approximately 22% for single-qubit errors and 18% for qubit-link errors. The model also exhibits strong ranking agreement, with the ordering induced by predicted error values closely matching that of the actual calibration errors, as reflected by high Spearman correlation. The framework consistently identifies weak links and high-noise qubits and remains robust under realistic temporal noise drift.
On the IBM Quantum ibm_fez chip, Nervous Machine identifies 17 clusters of qubits that move together across 106 days of calibration telemetry, 15 of them multi-event-confirmed via independent coordinated-motion events. 156 qubits × 106 days of calibration telemetry. The substrate independently discovered 17 shared-resource clusters from raw telemetry alone, no privileged coupling-map input. 15 of 17 are multi-event-confirmed — at least two distinct coordinated-motion events with within-cluster correlation magnitude |r| > 0.5 in independent windows. Zero discovered clusters are quiet; every one shows coordinated motion at least once in 106 days. The largest discovered cluster discovered_shared_readout_chain_8ef82ebd covers exactly the 9 qubits described in the How Uncertainty Evolved Around a 9-Qubit Cluster substack post, with two confirmed events (March readout-only, May combined readout+T1). Top 8 discovered clusters by event count: Cluster ID Type Size Members Events T1 co-moves Peak |r| 78e15a79 shared readout chain 2 Q14, Q85 4 1/4 0.999 1cee4ce7 shared readout chain 2 Q20, Q153 3 2/3 0.997 3625b1c4 shared readout chain 2 Q58, Q109 3 1/3 0.984 49b87d14 shared readout chain 2 Q81, Q142 3 0/3 0.975 8ef82ebd ★ shared readout chain 9 Q17, Q24, Q36, Q52, Q53, Q55, Q61, Q73, Q91 3 2/3 0.811 d7dd4ee8 regional readout 2 Q115, Q132 2 0/2 0.999 f441a7fb regional readout 2 Q6, Q8 2 0/2 0.994 72cad12f shared readout chain 3 Q1, Q7, Q135 2 0/2 0.993 ★ = the cluster from the original substack post NM vs comparator benchmarks Same comparators (spectral clustering on the correlation matrix; isolation forest and per-day z-score for the event-window AUC test) applied to the IBM_Fez analysis: Test NM Spectral best Quantum: bipartite-split recovery on the 9-qubit cluster (ARI) 1.000 1.000 Quantum: event-window detection AUC 0.718 0.369 (isolation forest), 0.344 (z-score) — below chance Per-component anomaly methods score below chance on event-window detection because the event manifests as correlated cross-component behavior, and per-component detectors don't see cross-component structure by construction. NM's rolling correlation-magnitude signal is structurally appropriate for the event class and reaches AUC 0.718. Bundle contents README.md — framing quantum/ — IBM_Fez analysis code, extracted intermediate JSON, all figures, comparator benchmark LICENSE — CC-BY-4.0 Key figures and reports (inside the bundle): quantum/results/figure_multi_event_topology.png — timeline of all 17 IBM_Fez clusters' coordinated events quantum/results/figure_event_timeline.png — IBM_Fez March + May event windows quantum/results/comparator_table.md — full quantum comparator breakdown (spectral, hierarchical, PC algorithm, isolation forest, per-day z-score) Datasets cited IBM_Fez calibration archive — proprietary; pulled via IBM Quantum API. Extracted intermediate JSON files included in quantum/ for replication without database access. Companion writeup Substack series at https://nervousmachine.substack.com/ — How Uncertainty Evolved Around a 9-Qubit Cluster (the original quantum finding). Ongoing agent cluster analysis and daily predictions: https://quantum-atlas.nervousmachine.com License CC-BY-4.0 — see LICENSE file in the bundle.
Quasiparticle poisoning following particle impacts poses a significant challenge to the development of fault-tolerant superconducting quantum computers, as a sudden excess of quasiparticles can simultaneously degrade the coherence of multiple qubits across large device arrays. In this work, we present a statistical analysis that models the time evolution of radiation-induced qubit energy relaxation through quasiparticle density dynamics. This study provides insight into quasiparticle loss processes by distinguishing between recombination and trapping decay channels and assessing their respective impact on qubit performance. We precisely measure quasiparticle recombination in multiple transmon qubits and uncover an unexpected dependence of qubit relaxation dynamics on deposited energy. By linking correlated relaxation events across qubits to ballistic phonon propagation, we introduce a statistical localization approach to extract the energy deposited in the substrate, which is in good agreement with Monte Carlo simulation. This work establishes the quantitative framework for using an arbitrary subset of superconducting transmon qubits in a QPU as energy-resolving witness particle detectors.
Abstract Pulse-level calibration metrics — coherence times T₁ and T₂ measured through single-qubit Rabi and Ramsey sequences — are the standard by which quantum hardware providers certify device health. We demonstrate, through 191 circuit executions on three 156-qubit IBM Heron r2 processors (ibm_kingston, ibm_fez, and ibm_marrakesh), that these pulse-level metrics can give opposite conclusions to circuit-level diagnostics on the same device at the same time. Specifically, between experimental sessions on July 3 and July 5, 2026, IBM reported ibm_marrakesh T₂ improved from 303 µs to 340 µs (a 12% improvement), while the K–R scaling exponent α measured from GHZ circuit scaling increased from 1.406 to 3.811 (a 171% increase in coherent error dominance) and GHZ n = 2 infidelity rose from 0.018 to 0.204 — a 920% increase in circuit error. We further demonstrate that Ramsey circuit-context T₂ on qubit 0 is approximately 9–30 µs, compared to IBM-reported pulse-level T₂ of 32–340 µs — a discrepancy of 3–11×. We introduce five experimental demonstrations showing K–R circuit-level diagnostics detect degradation invisible to T₁/T₂, provide higher temporal sensitivity, enable algorithm failure prediction, and confirm the theoretical prediction αₜ v ₜᶜʰ + αₑₜₕₒ = 2 within 7% on real hardware. These findings establish circuit-level K–R diagnostics as a necessary complement to pulse-level calibration for operational quantum computing.
Pulse-level calibration metrics — coherence times T₁ and T₂ measured through single-qubit Rabi and Ramsey sequences — are the standard by which quantum hardware providers certify device health. We demonstrate, through 191 circuit executions on three 156-qubit IBM Heron r2 processors (ibm_kingston, ibm_fez, and ibm_marrakesh), that these pulse-level metrics can give opposite conclusions to circuit-level diagnostics on the same device at the same time. Specifically, between experimental sessions on July 3 and July 5, 2026, IBM reported ibm_marrakesh T₂ improved from 303 μs to 340 μs (a 12% improvement), while the K–R scaling exponent α measured from GHZ circuit scaling increased from 1.406 to 3.811 (a 171% increase in coherent error dominance) and GHZ n=2 infidelity rose from 0.018 to 0.204 — a 920% increase in circuit error. We further demonstrate that Ramsey circuit-context T₂ on qubit 0 is approximately 9–30 μs, compared to IBM-reported pulse-level T₂ of 32–340 μs — a discrepancy of 3–11×. We introduce five experimental demonstrations showing K–R circuit-level diagnostics detect degradation invisible to T₁/T₂, provide higher temporal sensitivity, enable algorithm failure prediction, and confirm the theoretical prediction αₜᵥₜᶜʰ + αₑₜₕₒ = 2 within 7% on real hardware. These findings establish circuit-level K–R diagnostics as a necessary complement to pulse-level calibration for operational quantum computing.