All GatesLiveGateSafetyGateOptGateChemGateShotValidatorDriftRadarQRNGLive ReplayStacked PipelineMonitoring

QEC PIPELINE DEMO

Three-stage layered selection that keeps shots outside the calibration window out of your application

PIPELINE CONFIGURATION

PIPELINE STAGES

SHOTVALIDATOR

Redundant syndrome check

InOut
--
--

--

Errors Caught

--

Agreement

SAFETYGATE

Correlated error detection

InOut
--
--

--

Errors Caught

--

Suspicion

LIVEGATE

Real-time early abort

InOut
--
--

--

Errors Caught

--

Doom Rate

ShotValidator
SafetyGate
LiveGate
Each stage passes surviving shots forward

HOW THE PIPELINE WORKS

1. SHOT VALIDATOR

Each stabilizer measurement is repeated 3 times. If any copy disagrees, the entire shot is flagged as corrupted and discarded. First line of defense.

2. SAFETY GATE

Surviving shots are scored for correlated error signatures (crosstalk, leakage, cosmic rays). Shots above the suspicion threshold are vetoed.

3. LIVE GATE

Real-time session monitor tracks shot quality round-by-round. When the signal says recovery is unlikely, it flags the session and triggers early abort.

PIPELINE GUARANTEES

  • No circuit modifications, works on standard QEC outputs
  • Worst case: all shots pass through all stages unchanged
  • Each stage is independently toggleable via SDK configuration
  • Both raw and gated estimates always returned for full transparency
  • Pipeline never increases total cost, only reduces wasted spend

SDK INTEGRATION

from sdk import (
    ShotValidator, SafetyGateV3, ExecutionMonitor,
    ValidationLayout, MonitoringSession, AbortPolicy,
)

# Stage 1: ShotValidator
layout = ValidationLayout.for_repetition_code(distance=5, fan_out=3, n_rounds=4)
validator = ShotValidator(layout, enabled=True)

# Stage 2: SafetyGate
gate = SafetyGateV3.from_calibration("ibm_fez", distance=5)

# Stage 3: LiveGate (via ExecutionMonitor)
monitor = ExecutionMonitor(
    session=MonitoringSession(backend="ibm_fez"),
    abort_policy=AbortPolicy(doom_threshold=0.20),
)

# Run pipeline
with monitor.track() as session:
    for shot in hardware_shots:
        # Stage 1: validate shot integrity
        sv_result = validator.score_shot(shot)
        if not sv_result.clean:
            continue

        # Stage 2: check for correlated errors
        sg_result = gate.evaluate(shot)
        if sg_result.vetoed:
            continue

        # Stage 3: real-time quality monitoring
        session.record(shot)

    print(f"Pipeline efficiency: {session.efficiency:.1%}")
    print(f"Errors caught: {session.total_errors_caught}")

Full pipeline available in QubitBoost SDK v2.4.0, pip install qubitboost-sdk