报告题目：Verifiable quantum advantage with peaked circuits

报 告 人：Yuxuan Zhang

报告时间：2026年1月15日14:30

报告地点：物理楼W260

内容简介：Over a decade after its proposal, the idea of using quantum computers to sample hard distributions has remained a key path to demonstrating quantum advantage. Yet a severe drawback remains: the anti-concentrated distributions believed to be hard to classically simulate are typically hard to verify without exponential classical effort. In this series of work, we propose peaked circuit sampling, addressing this challenge by using circuits that are otherwise random but engineered to produce a pronounced “peak” on a specific computational-basis string, providing a direct verification handle while retaining complexity-theoretic hardness.

To this end, I will summarize three complementary results. First, we analyze explicit “random + peaking” architectures and quantify when nontrivial peakedness can (and cannot) be generated efficiently, motivating peaked circuits as a plausible route to verifiable advantage. ([arXiv][1]) Second, I present average hardness results for generic random peaked circuits, including circuit-complexity lower bounds and average-case #P-hardness of estimating peakedness to high precision, along with connections to BQP/QCMA regimes relevant for efficient verification. ([arXiv][2]) Third, I discuss heuristic ways of constructing large peaked circuits, which have been experimentally demonstrated on Quantinuum’s H2 processor. We show results with an empirical gap between quantum runtimes and leading classical simulation strategies, positioning peaked circuits as both a benchmarking tool and a concrete step toward verifiable quantum advantage. ([arXiv][3])

arXiv: [1] 2404.14493, [2] 2510.00132, [3] 2510.25838

报告人简介：Yuxuan Zhang is a postdoctoral fellow working with Dmitry Abanin’s group across Princeton and EPFL, with research interests at the intersection of quantum computing and artificial intelligence. He earned his Ph.D. from The University of Texas at Austin advised by Scott Aaronson and Andrew Potter. After his PhD he worked as a CQIQC Postdoctoral Fellow with joint appointments at the University of Toronto and the Vector Institute for Artificial Intelligence, collaborating closely with Yong-Baek Kim (UofT) and Juan Carrasquilla (ETH Zurich).