DoctorGPT: Advancing Healthcare with Specialized Large Language Models

With the widespread application of large language models (LLMs) in natural language processing tasks, their capabilities in understanding and responding to human instructions have significantly improved. However, due to the lack of specialized training in the medical field, existing LLMs exhibit notable limitations when handling healthcare-related tasks, including low diagnostic accuracy, inappropriate medication recommendations, and the generation of erroneous medical information. These issues not only restrict the application scope of LLMs in healthcare but also pose potential risks to user health and safety. To address these challenges, this study introduces doctorGPT, a high-quality LLM instance specifically designed for medical dialogue. By leveraging a large-scale pre-trained model based on Qwen and fine-tuning it with extensive high-quality medical Q&A datasets, doctorGPT effectively enhances its performance in medical scenarios. Additionally, to further optimize the quality and coherence of model outputs, we adopted reinforcement learning from human feedback (RLHF), enabling doctorGPT to provide richer and more human-like responses in multi-turn dialogues. Experimental results demonstrate that compared to other open-source LLMs, doctorGPT exhibits superior performance.