Quantum Benchmark Zoo

← All benchmarks

Mid-circuit measurement benchmarking

Component-level

IBM's randomized-benchmarking suite that quantifies the error a mid-circuit measurement adds to the measured qubit and the dephasing and crosstalk it inflicts on unmeasured spectator qubits.

Measures
Error per mid-circuit measurement, non-QND error under repeated measurement, and measurement-induced dephasing and crosstalk on spectator qubits, from compared RB decays
Introduced by
Govia, Jurcevic & colleagues (IBM Quantum) (2022)
Also known as
Mid-circuit measurement randomized benchmarking, Mid-circuit measurement RB, MCM-RB, mcm-rb suite, MCM benchmarking

Mid-circuit measurement benchmarking (the “mcm-rb suite” of Govia, Jurcevic and colleagues at IBM Quantum) was the first randomized-benchmarking-based benchmark for measurements performed partway through a circuit. Mid-circuit measurements (MCMs) are the primitive behind error correction and dynamic circuits, and terminal-readout metrics miss both of their failure modes: the extra error a measurement adds to the measured qubit, and the backaction (dephasing and crosstalk) it inflicts on unmeasured neighbors.

How it works

The suite compares exponential decays from three sequence types run on an ancilla and nearby spectator qubits. mcm-rb interleaves a mid-circuit measurement of the ancilla into standard Clifford RB sequences; delay-rb replaces each measurement with a delay of equal duration; and mcm-rep repeats measurements back-to-back to expose non-QND error. Comparing the mcm-rb and delay-rb decays isolates what the measurement itself adds, in the style of interleaved RB: an error per measurement for the measured qubit, subject to interleaved-RB-like systematic bounds, plus measurement-induced dephasing and crosstalk rates for each spectator. The original demonstration characterized 23 of the 27 qubits of ibm_peekskill (a Falcon R8 device), running the suite simultaneously on up to 17 qubits (five ancilla-plus-control sets).

Strengths and limitations

The suite extends RB’s SPAM-robust decay-comparison logic to a neglected primitive and yields actionable per-qubit and per-pair rates: component-level numbers, not a whole-processor score. The name is also used as a generic topic label: this entry anchors on the IBM suite, but distinct MCM benchmarks now exist, including Sandia and Quantinuum’s scalable many-qubit MCM RB (Nat. Commun. 16, 5761 (2025)), a generalized cycle benchmarking for MCMs (PRX Quantum 6, 010310 (2025)), and Pauli noise learning for MCMs (PRL 134, 020602 (2025)). It was demonstrated on superconducting hardware but is not platform-specific in principle, and it is distinct from the same group’s dynamic-circuits RB, which additionally benchmarks feedforward.

Notable results

Published as New J. Phys. 25, 123016 (2023), with a maintained reference implementation (mcm_rb.ipynb) in qiskit-device-benchmarking. The Sandia-led review of quantum-computer benchmarking (Nature Reviews Physics 7, 105 (2025)) names mid-circuit measurements among the fault-tolerance primitives that benchmarks must cover.

Key papers

Reference implementations

  • Randomized Benchmarking : Estimates average gate error from the exponential decay of survival probability over increasingly long random Clifford sequences, robust to state-preparation and measurement errors.
  • Interleaved randomized benchmarking : Estimates the error of one specific quantum gate by interleaving it between random Cliffords and comparing the resulting decay against a reference randomized benchmarking curve.