Track the frontier, from condensed matter to cosmology.
Search papers, labs, and topics across Lattice.
This paper introduces Syndrome-Based Offline Logical Error Mitigation (SOLEM), a novel framework for quantum error mitigation that leverages syndrome information post-execution without modifying the original quantum circuit. SOLEM achieves exponential reductions in runtime overhead, comparable to adaptive syndrome-aware methods, by treating syndrome-conditioned logical error channels as inherent components of error-corrected circuits. The framework enables the use of more powerful decoders, allows for the construction and optimization of multiple estimators from the same data, and simplifies logical error characterization, all while maintaining significant efficiency gains.
Quantum error mitigation can be performed entirely offline using syndrome data, achieving exponential overhead reductions without circuit modification.
Recent work has shown that syndrome information obtained during error-corrected quantum computation can be utilized for logical error mitigation (LEM), yielding exponential reductions in runtime overhead compared to external'' LEM methods, which do not make use of syndrome data. However, these syndrome-aware LEM methods (SALEM) generally require adaptive circuit operations based on real-time processing of syndromes. Here, we introduce syndrome-based offline logical error mitigation (SOLEM), a framework that utilizes syndrome data entirely in post-processing, without any modification to the given error-corrected circuit. In this sense, SOLEM is built into'' error-corrected circuits, and is in fact simpler to implement than external LEM protocols. The basic idea of SOLEM is to use naturally occurring syndrome-conditioned logical error channels in place of deliberate circuit modifications. We develop unbiased and heuristic estimators within this framework, including offline logical versions of probabilistic error cancellation (PEC) and zero-noise extrapolation (ZNE), as well as a global-regression estimator. SOLEM's fully offline nature leads to three significant advantages. First, SOLEM can post-process syndromes with a more powerful decoder than the one used in real time. Second, multiple estimators can be constructed from the same measured data and optimized using empirical statistical errors and related performance metrics. Third, logical error characterization is needed only in post-processing, allowing characterization data to be collected during circuit execution and reducing sensitivity to hardware drift. Remarkably, SOLEM estimators can achieve these advantages while attaining a runtime overhead which is exponentially lower than that of external LEM and is comparable to that of corresponding SALEM estimators.