Abstract: [Purpose/Significance] Large language models（LLM）are leading the information retrieval field to transition from simple document retrieval towards a new stage of comprehensively satisfying users' information needs. Reviewing and exploring the evolutionary trends and future developments of LLM during this transition is of great significance for theoretical innovation and practical application extension in the field of LLM-enabled information retrieval. [Design/Methodology] This research selects multiple leading conferences in the field of information retrieval and analyzes them from dual perspectives of thematic and knowledge association evolution. It deeply investigates how frontier technologies like LLM are driving the evolutionary development, knowledge reconstruction, and innovative applications in information retrieval, thereby revealing the future development directions of the field under the influence of LLM. [Findings/Conclusion] Driven by LLM, the research themes and knowledge structures in the field of information retrieval are undergoing evolution. At the research paradigm level, there is a focus on new concept of human-machine collaboration, technological ethics, and the paradigm integration brought about by enhanced user experience. At the technical level, there is an emphasis on efficient retrieval model architectures and workflow optimization of LLM, the collaborative development of lightweight language models and LLMs, as well as the open-source and equitable development of LLM. However, the application of LLM in information retrieval still faces challenges such as difficulties in effective technical evaluation, uncertainties regarding the reliability of generated content, and the high complexity of social applications. [Originality/Value] This research introduces fine-grained knowledge association networks into the evolutionary analysis framework, providing a multi-dimensional analytical perspective for research in technology-enabled fields. It also clarifies and reveals the evolutionary patterns of the information retrieval field from the data level, delineating the future development of the field.

Keywords: Large Language Model（LLM）, Information retrieval, Topic evolution, Retrieval-augmented generation