Quantum optimization against the classical state of the art
Quantum computing, quantum machine learning and quantum annealing appear on 115 recent papers, mostly in AI journals, and 33 of them concern optimization. Firms in finance and logistics run quantum pilots. Do the quantum results beat the best classical solvers on the same instances, or only the classical baselines the papers chose?
What prompted this
Since October 2025, 115 papers carry a quantum keyword (81 in AI journals, 15 in economics, 13 in operations research): "quantum machine learning" on 22, "quantum computing" on 14, "quantum neural networks" on 9, "hybrid quantum-classical models" on 8. Thirty-three of the 115 also carry an optimization, annealing, combinatorial, portfolio, scheduling, logistics or finance keyword; "quantum annealing" appears on 3 and "quantum optimization" on 2.
Questions
- On portfolio, routing and scheduling instances used in the quantum papers, how do the reported solutions compare with Gurobi, CPLEX or a tuned metaheuristic given the same time?
- Do the hybrid methods' gains come from the quantum part or from the classical pre- and post-processing?
- At what instance sizes, if any, do current devices beat classical solvers on energy use rather than solution quality?
Data that could help
The instances and results published in recent quantum optimization papers, and standard classical solvers.
Next step
Re-solve the benchmark instances of ten recent quantum optimization papers with a commercial solver and a tuned local search under the same time limit.