Test a circuit against stated noise assumptions
Noise Lab compares ideal distributions with genuine local noisy simulations and a bounded readout-mitigation method. It is designed for learning, development, and transparent what-if analysis—not for presenting synthetic results as device experiments.
Simulation modes
Exact density-matrix evolution supports up to five active qubits. Seeded trajectory simulation extends selected Pauli, depolarizing, gate, and readout models to six through eight qubits. Limits are displayed before a run.
Noise and mitigation
Depolarizing, flip, damping, gate, and readout controls alter the calculation. Readout mitigation uses an independent confusion model derived from supplied values and reports whether it improved, had negligible effect, or worsened similarity.
Break This Circuit
A bounded sweep locates where a selected simulated threshold is crossed. The result is conditional on this circuit, model, shot count, and threshold—not a universal hardware limit.
Method and reference
Readout mitigation uses bounded iterative Bayesian unfolding against a simulated independent confusion model. The readout-noise unfolding reference explains the method family; QAtlas does not use device calibration data or claim hardware mitigation performance.
Frequently asked questions
- Are these hardware results?
- No. Noise Lab does not contact real hardware and labels every output as simulated, estimated, assumed, or user supplied.
- Why might mitigation worsen a result?
- Mitigation can increase variance or be a poor fit for the selected assumptions. Noise Lab compares the computed result rather than declaring mitigation automatically successful.