量子技术：系统级性能与验证优先级
Quantum Technologies: System-Level Performance and Validation Priorities
AI总结：
本文通过跨领域分析明确量子技术性能需结合任务、边界等多维度，提出量子优势需在匹配条件下对比，为量子技术系统级能力验证提供框架。
AI中文摘要：
量子技术的性能指标并非仅由硬件决定，而是取决于所声明的任务、系统边界、归一化分母、不确定性或安全协议以及对比对象。我们开展了涵盖量子计算、模拟、通信、传感、时钟、随机数生成及其支撑技术的跨领域分析，该分析改变了若干代表性案例的实质性结论。针对本文分析的最长直接有限密钥量子密钥分发案例，当以完整采集时间而非传输活跃时间归一化时，相同最终密钥的速率相差1.91倍。在高级LIGO中，6.1dB的峰值量子噪声抑制与0.534的符合分析就绪分数共存，将探测器级增益与实际观测服务区分开来。在量子计算和光子采样领域，匹配可观测量、误差容限、损耗模型、样本数量、摊销方式及经典硬件可改变或反转已发表的交叉优势主张。跨领域来看，反复出现的限制因素包括相关误差、乘性接口损耗、热与非平衡占据、校准协方差、测量效率、制造良率及控制延迟。因此，仅当针对固定任务和边界，在匹配的精度、耗时、可用性及生命周期成本下，接受的输出优于最佳已记录替代方案时，才能确立量子优势。由此产生的框架明确了将组件记录转化为可复现系统能力所需的测量，可信进展定义为可复现的逻辑工作负载、经前瞻性验证的模拟、优于直接传输的中继器链路、长时间校准的传感器与时钟，以及具备可预测良率和可靠性的集成硬件。
英文摘要：
Performance claims in quantum technology are not properties of hardware alone. They are properties of a declared task, system boundary, normalization denominator, uncertainty or security convention, and comparator. We perform a cross-domain analysis spanning quantum computing, simulation, communication, sensing, clocks, randomness generation, and their enabling technologies. The analysis changes substantive conclusions in several representative cases. For the longest direct finite-key quantum key distribution case analyzed here, the same final key gives rates differing by a factor of $1.91$ when normalized by complete acquisition rather than transmission-active time. In Advanced LIGO, $6.1\,\mathrm{dB}$ peak quantum-noise reduction coexists with a $0.534$ coincident analysis-ready fraction, separating detector-level gain from delivered observing service. In quantum computation and photonic sampling, matching the observable, error tolerance, loss model, sample count, amortization, and classical hardware moves or reverses published crossover claims. Across domains, the recurring limits are correlated error, multiplicative interface loss, thermal and nonequilibrium occupation, calibration covariance, measurement efficiency, fabrication yield, and control latency. A quantum advantage is therefore established only for a fixed task and boundary when the accepted output outperforms the best documented alternative at matched accuracy, elapsed time, availability, and lifecycle cost. The resulting framework identifies the measurements required to convert component records into reproducible system capability. Credible progress is defined by reproduced logical workloads, prospectively validated simulations, repeater links outperforming direct transmission, long-duration calibrated sensors and clocks, integrated hardware with predictable yield and reliability.
