The paper introduces QTRL (Quantum-Train Reinforcement Learning), a method that leverages quantum neural networks (QNNs) to generate parameters for classical policy models in reinforcement learning tasks. This approach addresses the challenges of data encoding and the need for quantum hardware during inference, making it more practical and cost-efficient. The QTRL framework combines a QNN with a mapping model to convert quantum measurement probabilities into classical policy parameters, reducing the number of parameters required for the classical model. Experiments in the CartPole-v1 and MiniGrid-Empty-5x5-v0 environments demonstrate that QTRL can achieve comparable or superior performance with significantly fewer parameters, highlighting its potential for real-world applications such as autonomous driving. The method's independence from quantum hardware during inference and its compatibility with classical transfer learning and fine-tuning techniques further enhance its practicality and cost-efficiency.The paper introduces QTRL (Quantum-Train Reinforcement Learning), a method that leverages quantum neural networks (QNNs) to generate parameters for classical policy models in reinforcement learning tasks. This approach addresses the challenges of data encoding and the need for quantum hardware during inference, making it more practical and cost-efficient. The QTRL framework combines a QNN with a mapping model to convert quantum measurement probabilities into classical policy parameters, reducing the number of parameters required for the classical model. Experiments in the CartPole-v1 and MiniGrid-Empty-5x5-v0 environments demonstrate that QTRL can achieve comparable or superior performance with significantly fewer parameters, highlighting its potential for real-world applications such as autonomous driving. The method's independence from quantum hardware during inference and its compatibility with classical transfer learning and fine-tuning techniques further enhance its practicality and cost-efficiency.