Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
Epidemic Prevention and Control of Data Pre-train Large-language Model Application
1Campus Hospital, Bohai University, China.
2College of Information Science and Technology, Bohai University, China.
ABSTRACT
The transformer-based GPT series models developed by OpenAI, have achieved significant breakthroughs in artificial intelligence. Upon exploring the underlying principles, it is apparent that the transformer architecture, combined with more extensive and cleaner datasets, human data annotation, and reinforcement learning methods such as RLHF, can lead to qualitative improvements in the accuracy and fluency of text generation. Consequently, our study attempts to collect publicly available COVID-19 statistical data and integrate it with human-annotated reinforcement reward training methods. The aim is to fine-tune the GPT-2 model, thereby generating a reinforced model that could serve as a preventative measure against the next wave of the COVID-19 pandemic. This model promises superior intelligence in answering questions about the prevention, and control of the next wave of COVID-19, providing more detailed responses than generic models. Thus, we expect to obtain an enhanced version of a GPT-2 generative text dialogue model, fine-tuned for epidemic prevention and control.
Keywords: AI, Large-language model, Fine-tune, GPT, Epidemic prevention and control.
Download PDF
1Campus Hospital, Bohai University, China.
2College of Information Science and Technology, Bohai University, China.
ABSTRACT
The transformer-based GPT series models developed by OpenAI, have achieved significant breakthroughs in artificial intelligence. Upon exploring the underlying principles, it is apparent that the transformer architecture, combined with more extensive and cleaner datasets, human data annotation, and reinforcement learning methods such as RLHF, can lead to qualitative improvements in the accuracy and fluency of text generation. Consequently, our study attempts to collect publicly available COVID-19 statistical data and integrate it with human-annotated reinforcement reward training methods. The aim is to fine-tune the GPT-2 model, thereby generating a reinforced model that could serve as a preventative measure against the next wave of the COVID-19 pandemic. This model promises superior intelligence in answering questions about the prevention, and control of the next wave of COVID-19, providing more detailed responses than generic models. Thus, we expect to obtain an enhanced version of a GPT-2 generative text dialogue model, fine-tuned for epidemic prevention and control.
Keywords: AI, Large-language model, Fine-tune, GPT, Epidemic prevention and control.
Download PDF