Task-Oriented Gaussian Optimization for Non-Gaussian Resources in Continuous-Variable Quantum Computation

In continuous-variable systems, non-Gaussian resources are essential for achieving universal quantum computation that lies beyond classical simulation. Among the candidate states, the cubic phase state stands out as the simplest form of single-mode non-Gaussian resource, yet its experimental preparation still remains a great challenge. Although a variety of approximate schemes have been proposed to simulate the cubic phase state, they often fall short when deployed in concrete quantum tasks. In this work, we present a Gaussian optimization protocol that systematically refines the non-Gaussian resources, which significantly improves the performance of both magic-state-based and measurement-based quantum computation. Leveraging task-specific Gaussian operations on approximate cubic phase states, our protocol offers an experimentally feasible approach to enhance gate fidelity in magic-state-based quantum computation and reduce the variance of nonlinear quadrature measurement in measurement-based quantum computation. Building on this framework, we further propose a task-oriented non-Gaussian state preparation scheme based on superpositions in the Fock basis followed by squeezing and displacement. This strategy enables direct tailoring of resource states to specific task goals. Owing to its flexibility and generality, our framework provides a powerful and broadly applicable tool for enhancing performance across a wide range of continuous-variable quantum information protocols.