Error Mitigation of Fault-Tolerant Quantum Circuits with Soft Information

Quantum error mitigation (QEM) is typically viewed as a suite of practical techniques for today’s noisy intermediate-scale quantum devices, with limited relevance once fault-tolerant quantum computers become available. In this work, we challenge this conventional wisdom by showing that QEM can continue to provide substantial benefits in the era of quantum error correction (QEC), and in an even more efficient manner than it does on current devices. We introduce a framework for logical-level QEM that leverages soft information naturally produced by QEC decoders, requiring no additional data, hardware modifications, or runtime overhead beyond what QEC protocols already provide. Within this framework, we develop and analyze three logical-level QEM techniques: post-selection and runtime abort policies, probabilistic error cancellation, and zero-noise extrapolation. Our techniques reduce logical error rates by more than 100x while discarding fewer than 0.1% of shots; they also provide in situ characterization of logical channels for QEM protocols. As a proof of principle, we benchmark our approach using a surface-code architecture and two state-of-the-art decoders based on tensor-network contraction and minimum-weight perfect matching. We evaluate logical-level QEM on random Clifford circuits and molecular simulation algorithms and find that, compared to previous approaches relying on QEC only or QEC combined with QEM, we can achieve up to 87.4% spacetime overhead savings. Our results demonstrate that logical-level QEM with QEC decoder soft information can reliably improve logical performance, underscoring the efficiency and usefulness of QEM techniques for fault-tolerant quantum computers.

💡 Research Summary

The paper challenges the prevailing view that quantum error mitigation (QEM) is only useful for noisy intermediate‑scale quantum (NISQ) devices and becomes irrelevant once fault‑tolerant quantum computing (FTQC) with quantum error correction (QEC) is available. By exploiting the “soft information” naturally produced by modern QEC decoders, the authors construct a complete logical‑level QEM framework that requires no extra quantum experiments, hardware modifications, or runtime overhead beyond the standard QEC cycle.

Key Concepts

• Decoder soft information: Instead of outputting a single most‑likely error pattern (hard decision), a decoder can provide the posterior probabilities of each logical Pauli class (I, X, Y, Z) conditioned on the observed syndrome. This probability distribution contains all the information needed to characterize the logical noise channel.
• Unbiased, Rao‑Blackwellized estimator: By averaging the posterior probabilities over many syndrome measurements, the authors obtain an unbiased estimate of the logical channel with reduced variance, eliminating the need for separate gate‑set tomography or randomized benchmarking.

Three Logical‑Level QEM Techniques

1. Post‑selection and Runtime Abort – Shots whose posterior logical error probability exceeds a preset threshold are discarded or cause the algorithm to abort early. Simulations show >100× reduction in logical error rates while discarding <0.1 % of shots.
2. Probabilistic Error Cancellation (PEC) – Using the soft‑information‑derived logical channel, the authors construct a signed quasi‑probability decomposition of the inverse channel. An adaptive shot‑allocation strategy minimizes the sampling overhead, achieving the same error suppression as conventional PEC but with far fewer total shots.
3. Zero‑Noise Extrapolation (ZNE) – Soft information enables direct measurement of the logical channel at amplified noise levels (θ > 1). This yields accurate extrapolation to the zero‑noise limit without assuming smooth scaling, reducing the polynomial overhead traditionally associated with ZNE.

Implementation Details
Two state‑of‑the‑art decoders are employed for a planar surface‑code architecture:

• Tensor‑network decoder: Exact computation of posterior probabilities via a two‑dimensional tensor network with Kronecker‑delta constraints and diagonal tensors for local Pauli error probabilities. Exact contraction is #P‑complete; the authors use a matrix‑product‑state (MPS) approximation with bond dimension χ≈16‑32, achieving O(n χ³) runtime while preserving high accuracy.
• Boundary‑enforced Minimum‑Weight Perfect Matching (MWPM) decoder: A fast approximation that enforces boundary conditions to map logical classes to specific defect pairings. Though not maximum‑likelihood, it provides sufficiently accurate soft probabilities for real‑time use.

Numerical Evaluation
Simulations on random Clifford circuits and molecular‑simulation workloads (e.g., H₂ electronic structure) were performed for surface‑code distances d = 7–31. Results include:

• Logical error rate reductions of 100–200× across all test cases.
• Shot‑loss rates below 0.12 % for post‑selection.
• Space‑time overhead savings up to 87.4 % compared with a pure‑QEC architecture, and up to 65 % relative to a GST‑based QEC + QEM approach.
• PEC and ZNE pipelines required 5–10× fewer shots than conventional implementations while maintaining comparable accuracy.

Implications for FTQC
Because the soft‑information pipeline piggybacks on the existing QEC cycle, no additional quantum resources are needed. The approach enables adaptive error suppression, dynamic abort decisions, and in‑situ channel characterization, all of which can be integrated into future large‑scale quantum processors. The authors discuss extensions to non‑Pauli noise, multi‑logical‑qubit operations, and experimental validation on real hardware.

Conclusion
By turning the probabilistic output of QEC decoders into a resource for logical‑level error mitigation, the paper demonstrates that QEM remains highly relevant—and even more efficient—in the fault‑tolerant regime. This work establishes a practical pathway to combine QEC and QEM, promising substantial reductions in both logical error rates and overall resource consumption for scalable quantum computing.