Since the reform and opening-up, China has continuously enhanced its policy cognition and accumulated policy experience through top-down unbalanced empowerment, bottom-up public opinion absorption and point-based pilot promotion, and formed a policy experiment model with Chinese characteristics. In essence, the policy experiment with Chinese characteristics is a process of knowledge production. Factors such as system, organization, information technology and spatial distance constitute the internal driving force of knowledge production. Factors such as experts, local officials and stakeholders build learning networks around the experimental topics, exchange experimental information, exchange experimental experience and share experimental resources, continue to transform fragmented policy knowledge into institutionalized knowledge, and even form technical programs, ultimately contributing to the production and reproduction of policy knowledge. According to the differences in the dominance of the central government and the level of enthusiasm of local governments, the learning modes of policy experiments with Chinese characteristics can be divided into interaction mode, political mobilization mode, waiting and seeing mode and learning while doing mode. Different learning modes produce different policy knowledge. To explore the knowledge production and learning mode of policy experiment with Chinese characteristics is conducive to deepening the policy cognition of China’s reform and development, and has important reference significance for the interpretation of China’s institutional civilization.