Środowiskowe Seminarium z Informacji i Technologii Kwantowych

Claims of quantum advantage increasingly depend not only on asymptotic scaling,but also on how runtime is defined and which classical baselines are used. Inthis talk, I will discuss two recent works that revisit these issues fromcomplementary perspectives. First, we reconsider reported scaling advantages inapproximate QUBO optimization and show that replacing PT-ICM with a strongerclassical reference based on simulated bifurcation largely closes the previouslyreported quantum-classical gap, while also illustrating how conclusions drawnfrom small problem sizes can be misleading. Second, we introduce experimentallygrounded, end-to-end runtime definitions for digital and analogue quantumcomputing and use them to reassess recent advantage claims under more realisticbenchmarking assumptions. The main message is that credible quantum-classicalcomparisons must be full-stack. Runtime estimates should include relevantsystem-level overheads, performance should be measured with operationallymeaningful metrics, and quantum methods should be compared against strongclassical implementations rather than weak reference points.