Skipper: Improving the Reach and Fidelity of Quantum Annealers by Skipping Long Chains

Quantum Annealers (QAs), limited by their single-instruction operation model and qubit connectivity, rely on chaining multiple physical qubits to enhance connectivity, reducing effective capacity by up to 33x. Our observations reveal that: (a) the distribution of QAs' chain lengths has a heavy tail, with dominant chains including more qubits; and (b) about 25% of qubits remain unused due to isolation. We introduce Skipper that enhances QA capacity and fidelity by pruning dominant chains. Using 5761-qubit QAs, Skipper enables tackling up to 59% (Avg. 28%) larger problems and enhancing fidelity by up to 44% (Avg. 33%), when eleven and five chains are trimmed, respectively. Skipper allows up to eleven chain cuts, running about 2,000 quantum executables. Skipper-G is a greedy approach that bypasses sub-problems unlikely to contain global optima, requiring only 23 executables for eleven cuts. Skipper-G can boost QA fidelity by up to 41% (Avg. 29%) when cutting five chains (11 runs). The Python implementation of Skipper and Skipper-G are available at 10.5281/zenodo.10932899.