Benchmark circuits cut quantum error bias by up to 15 percent

Joseph Harris, Kevin Lively and Peter K. Schuhmacher of the German Aerospace Center (DLR) have developed a new approach to reduce bias in quantum error mitigation, achieving up to 15% fidelity improvements on 100-qubit circuits with up to 2000 entangling gates. The researchers constructed verifiable benchmark circuits mirroring the structure and noise profile of application circuits to benchmark and mitigate error mitigation bias, requiring only the circuit and hardware’s native gate set.

They also present two methods for generating these circuits, one hardware-agnostic and another specifically for IBM superconducting hardware with no gate overhead, alongside a new technique called benchmarked-noise zero-noise extrapolation. The team has released a software package to implement these error mitigation techniques.

Verifiable Benchmark Circuits Reduce QEM Bias Up To 15 Percent

This advance centers on the construction of verifiable benchmark circuits designed to mirror the noise characteristics of the circuits they are intended to refine. These circuits allow for the benchmarking and mitigation of bias inherent in QEM methods, requiring only the target application circuit and the hardware’s native gate set for implementation. The core of this technique lies in creating benchmark circuits that closely resemble the structure and noise profile of the application circuit, but remain classically verifiable.

This allows researchers to quantify and correct for biases in the error mitigation process itself, a critical step toward reliable quantum computation. The methodology introduces a “bias-mitigated” quantity derived from a Taylor series expansion, effectively reducing the impact of errors on the final results. Two distinct methods for generating these benchmark circuits are presented, offering flexibility for different hardware platforms.

One method is designed to function independently of the specific hardware, though it introduces a small overhead in single-qubit gates. The other method is tailored specifically for IBM superconducting hardware, achieving the same result without adding any additional gates. This hardware-specific optimization highlights the potential for further gains by aligning error mitigation strategies with the unique characteristics of individual quantum computing systems. While the benchmark circuits are designed to mimic the application circuit’s error behavior, the researchers acknowledge limitations in fully capturing all noise sources.

Specifically, the initial benchmark circuits do not introduce entanglement, potentially underestimating the impact of crosstalk noise. Despite this, the approach demonstrably reduces bias, bringing quantum computations closer to accurate and reliable results. “In many cases the bias is largely reduced but not entirely eliminated,” the researchers note, indicating an area for ongoing refinement.

As a practical application of this technique, the team introduced “benchmarked-noise zero-noise extrapolation” (bnZNE), an adaptation of the widely used zero-noise extrapolation (ZNE) method. This bnZNE approach, tested through both classical simulations and 100-qubit experiments on IBM superconducting hardware, consistently outperformed standard ZNE in measuring both single-site observables and two-site correlations along a one-dimensional qubit chain.

Hardware-Agnostic Benchmark Circuit Generation for Error Mitigation

These circuits enable a reduction in bias for error mitigation methods, demonstrated through analytical proof for zero-noise extrapolation under depolarizing noise, and verified via classical simulation. The other is specifically optimized for IBM superconducting hardware, eliminating that gate overhead entirely and potentially offering further gains through alignment of error mitigation strategies with specific hardware characteristics.

For an error mitigation method requiring N circuit runs, the bias-mitigated method requires only *2N* runs, a modest increase in computational expense that allows for potentially significant improvement in accuracy. This efficiency is important for scaling quantum computations to more complex problems.

The team suggests future research should focus on identifying benchmark circuit classes that accurately capture noise effects through both gate structure and entanglement, and on tailoring circuits to specific combinations of error mitigation method, application and hardware. This adaptation, along with the benchmark circuit generation techniques, offers a practical pathway toward more reliable quantum computations.

IBM Heron Architecture Enables Zero-Overhead Benchmark Circuits

The IBM Heron architecture allows for the construction of benchmark circuits that eliminate gate overhead during quantum error mitigation, a feat achieved by tailoring circuit generation to the specific characteristics of IBM’s superconducting hardware. The team’s approach is modular, allowing for the creation of different benchmark circuits tailored to specific applications, hardware platforms, or error mitigation methods.

Bias-Mitigated ZNE Outperforms Standard ZNE in Simulations

The fidelity of quantum computations saw improvements of up to 15 percent through a new bias-reduction technique applied to zero-noise extrapolation (ZNE), as demonstrated with circuits containing up to 2000 entangling gates. Researchers developed “benchmarked-noise zero-noise extrapolation” (bnZNE), a refinement of ZNE that uses carefully constructed benchmark circuits mirroring the noise characteristics of the target hardware. The team extended their analysis to the measurement of multi-qubit observables, specifically examining two-qubit correlations along a one-dimensional qubit chain.

The team verified their analytical evidence with classical simulation, demonstrating that the method effectively reduces bias in the presence of depolarizing noise. A software package was also provided for implementing the error mitigation techniques used in this research.

100-Qubit Experiments Validate Bias Reduction on ibm_fez

These results, achieved on 100-qubit circuits, represent an advance over previous demonstrations limited to circuits with at most six qubits. Analytical evidence confirms that the method reduces bias under depolarizing noise conditions, a common source of error in superconducting quantum computers. A software package implementing these techniques is also available, facilitating wider adoption and further research in the field.

Software Package Facilitates Implementation of Error Mitigation Techniques

A software package accompanies newly detailed error mitigation techniques, allowing wider access to methods that improve the accuracy of quantum computations. The package facilitates implementation of a process yielding up to a 15 percent fidelity improvement over standard quantum error mitigation on circuits containing 100 qubits.

This dual approach provides flexibility for researchers working with diverse quantum platforms. bnZNE builds upon recent adaptations of ZNE that aim to refine the classical extrapolation process by benchmarking error rates. The availability of the software package, alongside the numerical data generated during the research, is a key component of this work, and researchers can now readily implement and test these techniques, accelerating the development of more robust and reliable quantum algorithms.

👉 More information
🗞 Reducing quantum error mitigation bias using verifiable benchmark circuits
✍️ Joseph Harris, Kevin Lively and Peter K. Schuhmacher
🧠 DOI: http://link.aps.org/doi/10.1103/vmwl-qymb

Stay current

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

Avatar of Ivy Delaney

Ivy Delaney

Ivy Delaney has been working with neural networks and machine learning since the mid-nineties, back when a couple of hidden layers and a long afternoon of training counted as ambitious. She has watched the field go from academic curiosity to the thing quietly running underneath everything, and she brings that long view to quantum computing. For Quantum Zeitgeist she covers the ground where the two fields meet. That means quantum machine learning and the variational algorithms it leans on, and it also means the less glamorous but more interesting story of classical machine learning already doing real work inside quantum machines, decoding error-correcting codes, calibrating noisy hardware and learning the error models that simulators depend on. She writes about the hardware those algorithms have to run on too, and about the post-quantum cryptography scramble that the same hardware has set off. Her stories typically start with the paper, whether that is peer-reviewed work, conference proceedings or an arXiv preprint, with the source linked so you can hold a claim up against the research it came from. She is unimpressed by benchmarks that will not say what they beat, and by demonstrations that only work in the press release.

Latest Posts by Ivy Delaney: