We have developed a 20-qubit superconducting quantum processor featuring frequency-tunable qubits and adjustable coupling strengths to enhance single- and two-qubit gate fidelities. A rapid calibration framework using automated tuning and reinforcement learning can achieve a 99.6% two-qubit gate fidelity. Furthermore, we implemented an industrial-compatible, non-lift-off fabrication process that reduces surface contamination and significantly improves overall qubit performance and scalability.

This technology overcomes the bottlenecks of superconducting qubits through three key innovations:
    1.20-Qubit Tunable Architecture: Features frequency-tunable qubits and adjustable couplers to maximize single- and two-qubit gate fidelities. 2.Automated Calibration: Utilizes reinforcement learning to achieve high-fidelity two-qubit gate. 3.Standard Process: Adopts semiconductor fabrication processes to minimize organic contamination, enable scalable manufacturing, and improve qubit performance.

Taiwan is advancing next-generation computing via a CPU+GPU+QPU architecture. By integrating automated calibration and reinforcement learning, we enhance superconducting quantum computer performance. We established a standard fabrication process using industrial-grade equipment, improving qubit stability, uniformity, and coherence. This enables large-scale production of high-quality chips, laying the foundation for future mass production and commercialization.