Abstract: Artificial intelligence, represented by Large Language Models (LLMs), has been deeply integrated into all stages of scientific research, driving a paradigm shift toward AI for Science (AI4S). This transformation compels libraries, whose knowledge services have traditionally focused on resource organization and information access, to undergo a fundamental paradigm transition. Based on an analysis of the impacts of AI4S on knowledge production and service mechanisms, this study proposes a new paradigm of library research knowledge service covering the entire lifecycle of scientific research. A hierarchical model, comprising the data foundation layer, knowledge computing layer, intelligent agent layer, and collaborative ecosystem layer, is constructed to elaborate its operational logic, followed by the design of a systematic evolutionary path for libraries toward intelligent transformation. The AI4Soriented knowledge service paradigm can optimize the processes of scientific knowledge organization and enhance knowledge utilization efficiency. It further enables libraries to be deeply embedded in the research innovation process, reshaping their functional boundaries and value positioning within the academic ecosystem, and providing theoretical reference and practical insights for service transformation in the era of artificial intelligence. 

Key words: Artificial intelligence, AI4S, Research knowledge services, Service paradigm shift