Randomized benchmarking is a core protocol family in quantum computing for quantifying gate performance under realistic noise, while reducing sensitivity to state-preparation and measurement (SPAM) errors. It provides operational metrics that are directly useful for comparing devices, calibrating control stacks, and tracking hardware improvement over time.

Our work focuses on extending randomized benchmarking beyond idealized Markovian assumptions toward practical non-Markovian settings. The publications below establish a general framework for non-Markovian randomized benchmarking, develop operational Markovianization tools, and connect benchmarking observables with machine-learning approaches for characterizing average non-Markovianity. Together, these results strengthen the theoretical foundation and practical utility of benchmarking in near-term quantum platforms.