Research on Fine-Tuning and Prompt Engineering of LLMs for Scientific Literature in Power Grid

Facing the challenges of the exponential growth of power grid scientific literature and the inefficiency of traditional analysis methods, this paper proposes a framework for constructing a large language model (LLM) for power grid scientific literature based on fine-tuning and prompt engineering. Through multi-task learning and domain adaptation, the framework enhances the model's ability to understand professional terminology and logic. Experimental results show that this framework significantly improves the processing efficiency, accuracy, and user satisfaction of literature in power grid domain compared with traditional methods. This frame provides an efficient solution for knowledge mining in the power grid domain.