GRADE: Grover-based Benchmarking Toolkit for Assessing Quantum Hardware

TL;DR

GRADE proposes a Grover-based benchmarking framework to evaluate quantum hardware reliability by running Grover's search over a user-defined set of primitives. It introduces a dynamic multi-target oracle and a composite scoring function defined as $Score = P_T - \lambda \sigma_T - \mu P_N$, where $P_T$ is the cumulative probability of target states, $\sigma_T$ the dispersion of target probabilities, and $P_N$ the non-target probability; $\lambda$ and $\mu$ are weighting hyperparameters. The toolkit is open-source and validated across multiple simulators and hardware providers, demonstrating cross-platform applicability without hardware-specific optimizations. The work provides a baseline, extensible framework for identifying performance bottlenecks and guiding improvements in Grover-based quantum algorithm implementations on diverse architectures.

Abstract

Quantum computing holds the potential to provide speedups in solving complex problems that are currently difficult for classical computers. However, the realization of this potential is hindered by the issue of current hardware reliability, primarily due to noise and architectural imperfections. As quantum computing systems rapidly advance, there exists a need to create a generalizable benchmarking tool that can assess reliability across different hardware platforms. In this paper, we introduce GRADE (Grover-based Reliability Assessment for Device Evaluation), an open-source benchmarking toolkit to evaluate the reliability of quantum hardware using a generalized form of Grover's algorithm. GRADE operates by implementing Grover's algorithm to search through a variety of primitive collections that are customizable by the user, analyzing the probability distribution of the results to assess accuracy and stability. This approach aims to evaluate the hardware performance of Grover's algorithm, which is fundamental in unordered search problems, making it an ideal candidate for benchmarking purposes, as it is one of the few generalizable quantum algorithms that provide a direct speedup. Importantly, GRADE can adapt to a wide range of quantum computing platforms, ensuring it can be applied across different hardware architectures. Additionally, our work provides a scoring function for evaluating the hardware performance for multi-target Grover's implementation. Our work validates this approach across a multitude of simulators and hardware platforms from varying providers, demonstrating its adaptability to differing backend providers.

Figures (4)

• Figure 1: Heatmap comparison between different $\mu$ values, showcasing effect on GRADE scores.
• Figure 2: GRADE score of various fake backends, with $\lambda$ and $\mu$ set to 0.
• Figure 3: GRADE score report for fake Torino, benchmarked for 1 to 4 targets in search space with size 8, with $\lambda$ and $\mu$ set to 1.
• Figure 4: GRADE score report for IonQ Aria 1, Rigetti Ankaa 2, Quantinuum H1-1, IBM Sherbrooke, and IBM Kyiv, benchmarked for 1 target in search space with size 8, with $\lambda$ and $\mu$ set to 1.