A REPRODUCIBLE QUANTUM EXPERIMENT
A controller learns to quiet hidden quantum errors using only syndrome measurements. Explore the real experiment, one correction at a time.
BlueQubit CPU coherent simulation
Noise, syndrome measurement & recovery run locally.
Conceptual illustration · inspect the measured data below
01 / EXPERIMENT EXPLORER
Move through the first calibration. Each point is a measured SPSA candidate, using 4,000 fresh probe shots.
Deployment uses the averaged correction from the second half of SPSA. The replay shows individual candidates.
All six curves are drawn; the unencoded curve extends above the encoded-detail view.
| Method | Failures | Error | 95% interval |
|---|
One seed, one noise model. The 11× comparison describes the final observed counts, not a guaranteed improvement. Round readouts share simulated trajectories; uncertainty intervals are per point. No hardware performance or fault-tolerant threshold is claimed.
02 / THE IDEA
The optimizer sees syndromes.
The hidden field stays hidden.
Hidden rotations disturb a three-qubit memory. Stabilizer measurements provide a signal without supplying the logical state to the optimizer.
SPSA adjusts per-qubit correction angles to lower the finite-shot syndrome cost. True noise parameters and logical readouts are excluded from its interface.
Apply the learned inverse rotations alongside QEC recovery. Compare the observed logical failures against the baselines and oracle.
03 / OPEN EVIDENCE
The code, observations, provider job IDs, and report are public. Run the same experiment with your own environment.
View the repository ↗Local simulation runs without an API key.
vfqec run E1 --backend local