Exponential measurement error mitigation in quantum sampling

Abstract Measurement errors limit the fidelity of sampling-based quantum tasks, yet extending error mitigation to sampling remains challenging. Here, we introduce a simple and scalable repetition-based scheme that mitigates measurement errors directly at the sampling level. By encoding each data qubit into multiple physical qubits prior to readout and applying majority-vote decoding, physical errors are converted into a higher-order process, enabling exponential suppression that persists despite realistic gate imperfections during encoding. Using experimental noise profiles from leading superconducting platforms, we demonstrate that a modest linear qubit overhead suppresses measurement errors by several orders of magnitude, significantly boosting fidelities for typical sampling tasks, such as measurement-based quantum computation and random circuit sampling. These results establish a practical, hardware-compatible, and scalable approach for improving the reliability of near-term quantum sampling experiments.