The Signal
Study reports higher simulated Max-Cut ratios with cat-qubit QAOA
A Scientific Reports study found higher simulated approximation ratios for cat-qubit QAOA than for standard qubit encodings on random Max-Cut instances. The comparison used numerically simulated gate sequences and equal average gate fidelities; the paper also identifies experimental limitations in parts of its assumed gate set.

Simulated Max-Cut results
A numerical study published in Scientific Reports reports that cat-qubit implementations of the quantum approximate optimization algorithm (QAOA) produced higher approximation ratios than standard qubit encodings on simulated random Max-Cut instances. The researchers compared the approaches at equal average gate fidelities, simulating gate sequences on Kerr nonlinear resonators and benchmarking them against ideal two-level systems subject to single-photon loss.
The paper, published on 10 November 2026, describes numerical simulations rather than a demonstration of QAOA running on quantum hardware. For its larger Max-Cut benchmarks, the paper reports simulations of 6-, 8- and 10-qubit problems, with the cat-qubit approach yielding higher approximation ratios than the standard encoding. The authors also simulated a two-qubit Exact-Cover problem using the full master equation, describing it as a proof-of-principle demonstration.
How the encoding works
QAOA uses shallow quantum circuits to seek approximate solutions to combinatorial optimization problems, according to the paper. In the cat encoding studied, the logical states are represented by coherent states with opposite amplitudes, written as |α⟩ and |−α⟩. The paper says single-photon loss is the primary noise source in this encoding and creates a bias toward Z-type errors.
For larger simulated systems, the researchers used a Pauli transfer matrix formalism to characterize the error channel produced by single-photon losses in the computational subspace. The study links the performance improvement as a possibility to the noise bias and to the comparatively greater use of bias-preserving gates in its Max-Cut implementation. The authors describe that connection as a possible relation, rather than a settled explanation of the results.
Parameters and experimental limits
For the comparative Max-Cut analysis, the authors selected a coherent-state amplitude of α = 1.36, based on the value that gave the highest average gate fidelity in their analysis. The simulations used a single-photon loss rate of K/1500, which the paper says was within the order of magnitude of experimentally relevant values for Kerr-cat qubits. The paper says the fidelity optimum was relatively broad around α ≈ 1.3–1.4 for the loss rates considered.
The authors say the assumed average fidelity for the RX gate is somewhat higher than experimental values reported at the time of the paper. The page cites measured fidelities for resonant Kerr-cat qubits of 99.25% for Z(π/2) and 94.7% for X(π/2), and for detuned Kerr-cat qubits of 99.18% and 92.5%, respectively. The paper also states that no experimental implementation of a two-qubit RZZ gate had been reported for Kerr-cat qubits.
The authors say further improvements in non-bias-preserving gates are needed for a direct experimental realization of the full gate set used in their simulations. They describe the chosen operating regime as experimentally relevant, while noting that the assumed gate set is not yet directly demonstrated as a whole.