Latency-constrained Fidelity Optimization for NISQ Execution via Quantum Circuit Knitting

Executing quantum circuits on NISQ devices poses challenges due to noise, decoherence, and limited qubit counts. Circuit Knitting (CK) allows solving these issues by breaking down large circuits into smaller sub-circuits. However, the subsequent reconstruction adds sampling and processing time. Quantum Error Mitigation (QEM) can enhance output quality, but it also raises execution costs. To address the above issues, we introduce MOSAIQ, a cost-aware framework optimizing both circuit partitioning and error mitigation for reliable NISQ execution. MOSAIQ uses calibration-driven noise estimation, mitigation-aware fidelity modeling, and cut-dependent recon- struction costs to choose partitioning strategies that balance end- to-end fidelity with practical feasibility. We define the problem as an optimization task and test the MOSAIQ on noisy simulators and actual quantum HW. Results indicate that MOSAIQ prevents overly aggressive partitioning and improves the balance between fidelity, sampling overhead, and execution time.