A study of large language model Q&A based on LoRA Fine-Tuning and prompt engineering
Shuo Li, Ning Ma · 2025
Intelligent Q&A has been an important research direction in the field of natural language processing, and its research paradigm has undergone a significant change with the advent of the era of big language modeling. In this study, we adopt two models, Llama3-8B-Chinese-Chat and DeepSeek-llm-7B-Chat, as the base models, and use the MedQA dataset as the benchmark dataset, with the help of LoRA fine-tuning as well as cueing engineering and other techniques, to construct a model evaluation and optimization framework. The framework evaluates the intelligent Q&A ability of the big language model in multiple dimensions through scientific and rigorous methods, and further optimizes the two models under the premise of guaranteeing the computational efficiency of the models, so as to enhance the performance of the big language model in the field of intelligent Q&A in a targeted way, and to improve the logic, accuracy, and adaptability of the model in the specific task of Q&A. Finally, this study compares the overall performance of the two models before and after fine-tuning respectively. The results show that the accuracy and performance of the models in the intelligent Q&A task can be significantly improved by incorporating LoRA fine-tuning techniques and cue engineering strategies. The research results provide innovative theoretical ideas and practical basis for efficient reasoning and optimizing the fine-tuning process of large language models.