arXiv:2610.29035
Optimization versus fair sampling in quantum annealing: An exactly characterized frustrated Ising benchmark
Masaru Hitomi, Masayuki Ohzeki
Abstract
Reaching a degenerate ground-state manifold and sampling it uniformly are distinct properties of a quantum annealer. We quantify this distinction using an Ising ring with additional couplings between maximally separated sites. For antiferromagnetic benchmark couplings and $N \geq 5$, we derive the exact ground-state energy, degeneracy, and defect structure for each congruence class of $N$ modulo four. This classification separates ground-state hit probability from conditional sampling bias. We embed instances with up to 162 logical spins on a D-Wave Advantage processor using length-two chains and compare its output with simulated annealing on the embedded and logical problems. For a frustrated $N=83$ instance, the processor visits 164 of the 166 exact ground states, yet the conditional distributions over states and defect sectors remain nonuniform. For $N=160$, the processor has a larger ground-state hit probability than logical simulated annealing (0.869 versus 0.519), but also a larger total-variation distance from uniformity (0.317 versus 0.136). A pause during the anneal raises the ground-state hit probability from 0.42 to 0.86 without restoring uniform sampling. These results identify a relaxation window and demonstrate that optimization success and sampling quality need not improve together, providing an exactly characterized benchmark for open-system quantum annealing.
AI Summary
The paper asks how ground-state optimization success relates to uniform sampling of degenerate solutions in quantum annealing. It studies an antiferromagnetic Ising ring with additional maximally separated couplings, whose complete ground-state manifold is derived using bond-variable parity constraints and defect counting. For $N\geq5$, the degeneracy is $N$, $2N$, $2$, and $2N$ for $N\bmod4=0,1,2,3$, respectively. The classification enables separate measurements of ground-state hit probability $P_G$ and conditional total-variation distance from uniformity, including distributions over defect sectors. Instances up to 162 logical spins are implemented on a D-Wave Advantage processor using length-two chains and compared with simulated annealing on both embedded and logical problems. For $N=83$, a $20\,\mu\mathrm{s}$ anneal gives $P_G=0.394$, visiting 164 of 166 ground states, but state- and sector-level distances remain 0.300 and 0.165. For $N=160$, the QPU has higher $P_G$ than logical simulated annealing (0.869 versus 0.519) but greater state-level bias (0.317 versus 0.136). A $100\,\mu\mathrm{s}$ pause near annealing fraction $s=0.40$ raises $P_G$ from 0.422 to 0.862 without improving uniformity. The pause dependence identifies an intermediate relaxation window, interpreted as consistent with thermally assisted annealing. The data do not uniquely distinguish freeze-out, analog-control errors, and embedding effects as sources of sampling bias; further gauge, relabeling, and embedding controls are needed. The classical comparisons concern the specified protocols, not general classical hardness, and sparse ground-state counts complicate interpretation of simulated-annealing sampling distances.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 12 pages, 4 figures