VFQEC E1 - aca95affb4e946598029f8c307c01b80
BlueQubit cpu coherent simulation + local noise/syndrome/recovery
| Method | Final error | 95% interval | Fitted / round |
| Unencoded qubit | 0.25725 | [0.2439, 0.271] | 0.0873256 |
| Standard QEC | 0.0055 | [0.003635, 0.008314] | 9.4078e-05 |
| QEC + twirling | 0.004 | [0.002464, 0.006488] | 8.59271e-05 |
| QEC + calibrated field | 0.0005 | [0.0001371, 0.001821] | 1.26067e-05 |
| QEC + adaptive field | 0.0005 | [0.0001371, 0.001821] | 1.02507e-05 |
| Oracle (single-qubit field) | 0.0005 | [0.0001371, 0.001821] | 9.9054e-06 |
- Phenomenological data noise; ideal gates, preparation and terminal readout.
- Spatial decoding each round; q>0 is not fault-tolerant spacetime decoding.
- Calibration uses fresh encoded probes; drift time freezes during calibration.
- Oracle cancels single-qubit fields only; ZZ and stochastic noise remain.
- Endpoint curves share simulated trajectories and have correlated time points.
- Zero observed failures have nonzero confidence upper bounds.
- Rates are descriptive; coherent/drifting noise need not be exponential. No fit covariance is interpreted as independent-round uncertainty.
- unencoded: exponential model poorly describes coherent/drifting memory
