Learning from quantum random circuit sampling? A new benchmarking method.

Tudor Manole at the Statistics and Data Science Center, Massachusetts Institute of Technology, led a team that surpasses benchmarking limits in quantum random circuit sampling. Building on this work, Daniel K. Mark, Wenjie Gong, Bingtian Ye, and Soonwon Choi, all from the Center for Theoretical Physics, a Leinweber Institute at MIT, developed new analytical tools for assessing quantum computational complexity. Published in arXiv, their findings, co-authored with Yury Polyanskiy from both the Statistics and Data Science Center and the Department of EECS at MIT, offer deeper insights into the capabilities and limitations of near-term quantum devices.

This research establishes a more rigorous framework for evaluating quantum algorithms and advancing the field of quantum information science.

Advancing Quantum Benchmarking with Random Circuit Sampling Researchers at the Statistics and Data Science Center at MIT and the Center for Theoretical Physics, a Leinweber Institute, have introduced new benchmarking methods based on random circuit sampling (RCS) to substantially extend conventional approaches for characterizing quantum devices. According to Tudor Manole and his colleagues, this framework extracts rich diagnostic information, including spatiotemporal error profiles and correlated errors, without requiring modifications to existing experiments. These advancements address a critical need for accurate, in situ characterization of large-scale quantum systems, which are often impossible to simulate on classical computers due to numerous error sources fully. Building on this, the team, including Wenjie Gong, Bingtian Ye, and Yury Polyanskiy, developed techniques that avoid classically intractable simulations by leveraging side information obtained from reference quantum devices. Their approach centers on advanced high-dimensional statistical modeling of RCS data, allowing for characterization of information-theoretic limits of error estimation. Specifically, they derived matching upper and lower bounds on sample complexity across varying regimes of side information, identifying surprising phase transitions in learnability as more side information becomes available. Furthermore, Soonwon Choi and the research group demonstrated their methods using publicly available RCS data from a state-of-the-art superconducting processor. The resulting in situ characterizations were qualitatively consistent with component-level calibrations, yet quantitatively distinct, establishing both practical benchmarking protocols for current and future quantum computers. These findings also define fundamental information-theoretic limits on how much can be learned from RCS data, representing a significant step forward in the field of quantum error mitigation and device validation.

Decoding Quantum Errors Through Data-Driven Analysis Building on this advancement in benchmarking, Tudor Manole and colleagues from the Statistics and Data Science Center at MIT developed techniques to characterize error profiles without relying on classical simulations of the quantum circuit. Their approach leverages side information, bitstring samples from reference quantum devices, to achieve this, enabling analysis of spatiotemporal errors and correlated errors within the system. This method avoids the computational bottlenecks inherent in simulating large-scale quantum systems, opening new avenues for in situ characterization of quantum processors. Furthermore, Wenjie Gong, Bingtian Ye, and Soonwon Choi investigated the information-theoretic limits of error estimation, deriving matching upper and lower bounds on the sample complexity required for accurate characterization.

The team identified surprising phase transitions in learnability as the amount of side information varies, suggesting that there are optimal levels of reference data for maximizing the efficiency of error estimation. These findings, according to the research, provide a fundamental understanding of the trade-offs between data acquisition and accuracy in quantum error analysis. Yury Polyanskiy and the team demonstrated the effectiveness of their methods using publicly available random circuit sampling (RCS) data from a superconducting processor. The resulting in situ characterizations were qualitatively consistent with component-level calibrations, yet revealed quantitatively distinct insights into the processor’s error landscape. This suggests that the developed framework can provide a more comprehensive and nuanced understanding of quantum device performance than traditional calibration methods, establishing practical benchmarking protocols for both current and future quantum computers. Building on this advancement in benchmarking techniques, researchers at the Massachusetts Institute of Technology, including those from the Statistics and Data Science Center and the Center for Theoretical Physics, have unlocked a pathway to more thoroughly characterize quantum devices. This detailed analysis, extracting spatiotemporal error profiles and identifying contextual errors, could enable the development of more robust and reliable quantum systems. The implications extend beyond simple performance metrics, offering insights into error mitigation strategies crucial for scaling quantum technologies. For industries reliant on complex computations and secure data transmission, this represents a significant step toward realizing the full potential of quantum computing. By leveraging side information from bitstring samples, the MIT-CTP team avoids classically intractable simulations, paving the way for practical, in situ characterization of increasingly large and complex quantum devices. Source: https://arxiv.org/pdf/2510.09919 Tags: