一句话：通过运行时监控、性能预测与噪声缓解，提升 NISQ 上量子程序预言机测试的可观测性与可靠性。
该博士论文针对量子软件开发中的两大难题——概率语义与受限可观测性导致的理解困难，以及 NISQ 设备上执行可靠性不足——提出改进方案。工作分为三部分：（i）通过自动化运行时监控提升执行期间的可观测性；（ii）预测模拟器与量子硬件上的执行开销与性能波动；（iii）利用电路结构与后端上下文过滤含噪的结果分布，从而稳定基于预言机（oracle）的测试。未来工作将把电路切割与后处理缓解相结合：将程序分解为子电路、分别做噪声缓解，再重建最终输出分布，以将测试扩展到更大规模电路；并进一步探索概率误差消除（PEC），评估逐子电路缓解之外的精度—成本权衡。
Modern quantum computing platforms have moved from laboratory prototypes to cloud-accessible systems, but quantum software development still faces two fundamental challenges: limited program comprehension under probabilistic semantics and restricted observability, and limited execution reliability on noisy intermediate-scale quantum (NISQ) devices. This thesis addresses these challenges by improving observability through automated runtime monitoring and performance estimation, and by improving testing reliability through noise mitigation. Concretely, this thesis develops methods to (i) enhance runtime observability during execution, (ii) predict execution cost and performance variability on simulators and quantum hardware, and (iii) stabilize oracle-based testing by filtering noisy outcome distributions using circuit structure and backend context. As future work, this thesis will integrate circuit cutting with post-processing mitigation to scale testing to larger circuits by decomposing programs into subcircuits, mitigating them individually, and reconstructing the final output distribution, and will further explore probabilistic error cancellation (PEC) to assess accuracy-cost trade-offs beyond per-subcircuit mitigation.