Stale snapshots, not hidden physics: readout faults on three IBM Heron r2 processors are rare, persist for tens of minutes, and do not depend on measurement context (TESSERACT-RO-06/RO-07)
Two preregistered experiments (doi:10.5281/zenodo.24407897, doi:10.5281/zenodo.24419327) on ibm_marrakesh, ibm_fez and ibm_kingston; 27 jobs, 193 s QPU. RO-06, 8 rounds of an all-qubit readout probe: readout faults hidden from IBM's calibration (probe > 15%, IBM < 3%) were rare (H1 not supported); every such qubit recovered in another round (H2 supported); faults persist across 15-minute rounds 87% of the time. RO-07: readout error did not depend on context (isolated vs all-qubit readout; other qubits in 0 vs 1): all effects below 1.7 pp; both context hypotheses failed; controls clean. Exploratory: IBM's readout values did not change during 1.75 h of sampling, ibm_kingston's not for >15 h; when ibm_fez updated, the new values matched our latest probes (rho 0.95 vs 0.81 for the old values). The gap between IBM's numbers and in-job readout is stale calibration plus hour-scale qubit changes. Record contains paper, source, all counts, calibrations, locks, QPYs, analyses and code.
Authors
- Devin Phillip Davis
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-20
- DOI
- https://doi.org/10.5281/zenodo.24420901
- Primary Topic
- Quantum Computing Algorithms and Architecture
- Type
- preprint