All GatesLiveGateSafetyGateChemGateShotValidatorDriftRadarQRNGLive ReplayStacked PipelineMonitoring

LIVE GATE DEMO

Detect runs heading for failure in real time and stop them before the remaining shots are spent

Works across any quantum workload: QEC, VQE, QAOA, calibration, tomography, benchmarking.

SCENARIO CONFIGURATION

Savings scenario: Failure-risk score (35%) exceeds threshold (20%). Early abort will trigger, reducing spend on the remaining shots.

WITHOUT QUBITBOOST

Shots Executed0 / 100,000

,

Shots Run

,

Failure-Risk Score

,

Total Cost

,

Usable Shots

WITH QUBITBOOST

Shots Executed0 / 100,000

,

Shots Run

,

Failure-Risk Detected

,

Total Cost

,

Usable Shots

SAFETY & OPERATIONAL GUARANTEES

  • Live Gate never increases total cost
  • Worst-case behavior: no abort (full job runs as normal)
  • Abort decisions are statistically justified (confidence-based)
  • No changes to circuits, hardware, or user workloads
  • Fully optional and policy-controlled
  • Applicable to any executable quantum job, not limited to QEC workloads

What does "abort" mean?

An abort stops further execution when early evidence shows the job is statistically likely to fail. Partial results are preserved for diagnostics, but remaining shots are not run to avoid wasted cost.

HOW LIVE GATE WORKS

1. PROBE

Run an initial probe phase (~10% of budget) using probe circuits or workload-specific failure detectors.

2. EVALUATE

Calculate a statistical confidence interval on the failure-risk score. If the confidence lower bound exceeds the threshold, abort is triggered.

3. SAVE

Early abort prevents paying for shots the signal says will not recover. Savings depend on workload and abort timing.

HARDWARE EVIDENCE

LiveGate's failure detection is validated through ShotValidator hardware experiments on IBM Fez: 27 QEC configurations, 216K shots, 74% mean error capture. The hardest config (d=7, r=8, f=3) caught all 63 corrupted shots.

216K

Shots validated

74%

Mean error capture

63 → 0

Errors eliminated (d7)

View full hardware results on ShotValidator page →

FAQ

Could this abort a job that would eventually succeed?

Yes. Live Gate is a conservative safety layer. It may abort borderline jobs to protect budget when early evidence indicates elevated failure risk.

What if the failure-risk estimate is wrong early on?

Abort decisions are based on a statistical confidence interval, not point estimates. Abort only triggers when the lower bound of the confidence interval exceeds the configured threshold.

Is this backend- or workload-specific?

Live Gate uses backend-calibrated failure detectors but applies uniformly across workloads. Policy inputs are tunable per backend, cost tier, or experiment class.

Technical Notes

Live Gate performs per-shot failure detection during an initial probe phase using probe circuits. As shots accumulate, a statistical confidence interval is computed over the observed failure-risk score. If the lower bound of this interval exceeds the configured abort threshold, execution halts early to prevent further budget waste.