DLR, Cologne team reports state-loading method cutting two-qubit gate depth up to 100x
Researchers at the German Aerospace Center (DLR) Institute of Software Technology and the University of Cologne posted an arXiv paper, "Scalable Unitary State-Encoding for Generic Large Data," describing a deterministic state-preparation framework (the article does not say whether it has been peer reviewed, and arXiv papers are typically preprints). It combines tensor-network representations, Boolean logic minimization and Walsh-Hadamard transformations, and the authors report up to two orders of magnitude lower two-qubit gate depth than standard multiplexer compilers. In benchmarks on Lorentzian, polynomial and financial jump-diffusion distributions, depth plateaued at a fixed infidelity of about 5 x 10^-3, versus exponential scaling for Qiskit state-preparation tools. On the 20-physical-qubit IQM Emerald superconducting processor, a 2D Gaussian state was encoded at total circuit depth 56 (CZ depth 26) with state infidelity of 0.011 ± 0.002 over 256,000 shots.