Abstract: A 2-input 16 output quantum phase neural network (QPNN) is designed to implement 16 types of logic operators and is verified by the parameterized quantum circuits. The proposed QPNN comprises two real-valued inputs, a quantum state phase encoding function and 16 quantum logic operator units. Each quantum logic operator unit is realized through a parameterized quantum circuit with two input qubits and two auxiliary qubits. The circuit employs only five adjustable weight phases, which utilizes phase-tunable R_y gates for phase rotation, a controlled-R_y gate, and a Toffoli gate to achieve multi-qubit phase superposition. The real-valued output of the logic operator is derived by computing the square of the sine function value of the output phase, which corresponds to the probability of the quantum state in \left| 1 \right\rangle , thereby enabling any of the 16 logical operations. This paper also develops a learning algorithm with adjustable weight phases based on the gradient descent method and adaptive learning rate. The proposed work is implemented and validated on the parameterized quantum circuits constructed by using the Qiskit platform and compared with other networks in terms of structural complexity and implementation accuracy. This work opens up new possibilities for the practical application of multi-input/multi-output quantum phase neural networks, particularly through the Qiskit-based implementation of parameterized quantum circuits.