Software Technology Team shows quantum error mitigation may beat existing methods on IBM hardware

Quantum processors require error rates far below current capabilities to scale beyond today’s limits; randomized benchmarked-noise zero-noise extrapolation, bnZNE, is an improved quantum error mitigation technique that addresses this challenge. This low-cost approach consistently delivers more accurate results than standard zero-noise extrapolation when tested on both and IQM computer platforms across multiple applications while maintaining equivalent computational cost. Randomized bnZNE delivers more reliable results from existing quantum computers without increasing computing demands.

The improvement stems from better measurement of errors during calculations, offering a practical set of tools to achieve useful outcomes with today’s imperfect devices. Quantum computers promise transformative calculations but remain susceptible to errors; overcoming these limitations is key for scaling up current capabilities.

Randomized bnZNE, benchmarked-noise zero-noise extrapolation, is an improved quantum error mitigation technique designed to deliver more accurate results than standard methods without increasing computational demands. Like calibrating scales to ensure accuracy, readout error mitigation corrects imperfections when reading out final answers from a quantum computer. The team tested this new method on machines built by both and IQM, consistently observing improvements across multiple applications; but how effectively does it address deeper sources of noise within complex algorithms.

Randomized bnZNE enhances expectation value estimation for complex quantum algorithms

Randomized benchmarked-noise zero-noise extrapolation (bnZNE) delivers more accurate results than standard ZNE with large scale circuits, a significant leap forward from previous methods limited by accumulating errors in deep algorithms. The team’s innovation lies in randomized bnZNE, an enhanced technique that uses classically simulable “benchmark” circuits mirroring application circuit structure to sensitively measure effective error rates at each noise level.

Incorporating techniques like Pauli twirling and readout error mitigation alongside multiple random circuits per noise level further boosts accuracy without increasing computational cost compared to traditional approaches. Researchers at the Institute of Software Technology developed new ways to generate tailored benchmark circuits for IQM superconducting devices; this allowed sensitive measurement of effective error rates at scaled noise levels beyond integer values used in simpler extrapolation.

Benchmarking Noise Extrapolation with Structurally Similar Circuits

Randomized bnZNE refines how artificially added noise levels are interpreted, improving estimates of results as if no errors existed, a step beyond standard ZNE’s simple backward extrapolation from noisy data. While promising for more reliable quantum computing today, the team acknowledges limitations in exceptionally deep circuits; low signal-to-noise ratios can undermine fitting accuracy, a challenge inherent in all such “low cost” mitigation pushed too far.

This raises questions about whether improvements will plateau as algorithms become increasingly complex and noisy, potentially necessitating fundamentally different approaches alongside full error correction for fault tolerance. Distinguishing between true signals and accumulated errors becomes difficult at high circuit depths, making performance particularly sensitive. The Institute of Software Technology’s randomized bnZNE refines existing quantum error mitigation by improving interpretation of artificially added noise during calculations, allowing for more accurate estimations without errors, an important step towards practical applications. By employing classically simulated circuits mirroring the structure of complex algorithms, researchers sensitively measure effective error rates beyond simple scaling values used in standard techniques like Zero Noise Extrapolation (ZNE). This advancement consistently improves accuracy across multiple large-scale problems while maintaining equivalent computational cost; it is immediately deployable on hardware such as IQM systems without substantial investment in error correction infrastructure.

This research demonstrated benchmarked-noise zero-noise extrapolation (bnZNE), a method that refines how artificial noise levels are interpreted to improve estimates of quantum computation results. bnZNE builds upon existing methods by using classical simulations of application circuit structures to more accurately measure errors at each scaled noise level, offering improved performance compared to standard ZNE with the same computational cost. The team tested this approach successfully on IBM and IQM hardware for large-scale applications, showing its versatility as an alternative mitigation technique. Researchers observed limitations in very deep circuits where signal can be lost within accumulated error; they suggest further work may eventually require fundamentally different approaches alongside full error correction.

👉 More information
🗞 Improved Low-Cost Quantum Error Mitigation with Benchmarked-Noise Zero-Noise Extrapolation (bnZNE)
✍️ Joseph Harris, Kevin Lively and Peter Ken Schuhmacher (Institute of Software Technology)
🧠 ArXiv: https://arxiv.org/abs/2611.16971

Stay current

See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.

Avatar of Muhammad Rohail T.

Latest Posts by Muhammad Rohail T.: