Skip to content
RESEARCH · ERROR MITIGATION

DLR tests a better noise yardstick on IBM and IQM quantum chips

A new DLR preprint broadens tests of benchmarked-noise extrapolation, with the same total measurement budget and some wider uncertainty.

Published Source report Sources checked
Reading tools for text size, copying, printing and your reading list
My reading list

Researchers at the German Aerospace Center (DLR) report that a benchmark-based approach brought average quantum-computing estimates closer to known answers in tests on IBM and IQM machines. Their preprint, submitted on November 12, extends an existing error-mitigation method. The improvement came with greater variability in some tests.[1][2]

Error mitigation adjusts estimates from repeated noisy runs; it does not create error-corrected . Standard zero-noise extrapolation deliberately runs noisier versions of a circuit, the sequence of operations in a calculation, then estimates the answer at zero noise. Benchmarked-noise ZNE, or bnZNE, adds similarly structured circuits with predictable ideal answers. Their measured errors help set the scale used for that extrapolation.[3][4]

Joseph Harris, Kevin Lively and Peter K. Schuhmacher introduced bnZNE in March, already testing 100-qubit IBM circuits. The new work sets out a version that averages randomized circuit variants and adds an IQM-specific benchmark recipe. Experiments include simulations of interacting spins using 100 qubits on IBM Fez and 30 on IQM Emerald, plus smaller IBM Fourier-transform and search circuits.[5][6]

For each application, the authors allocated each method 196,608 shots, or repeated circuit measurements. bnZNE used twice as many circuits with half as many shots per circuit. They report better average agreement with reference answers. That comparison establishes an equal shot budget, not equal runtime or cloud bills.[6][2]

The limits matter. Longer IBM circuits showed substantial statistical spread. Seven high-uncertainty bnZNE points were excluded from a separate analysis of correlations between qubits, and the IQM analysis used quadratic extrapolation after comparing three fitting choices. The reported gains concern selected quantities; they do not demonstrate error-corrected computing or an advantage over classical computers.[2]

Keep your reading trail

READ NEXT

Continue exploring

All Research stories
  1. Whole-chip entanglement

    IQM reports entanglement across its full 54-qubit chip

    IQM reports a 54-qubit entangled state with measurement correction. Here is what its 0.516 fidelity says about the processor.

    Published · Source report · Sources checked · About 2 min read