SCD: Strengthen Character Development in Long Stories with LLMs via Prompt Engineering

Although large language models (LLMs) have achieved remarkable results in the field of natural language processing, they still exhibit shortcomings in character portrayal when generating long stories. To address this issue, this paper proposes SCD (Strengthen Character Development) framework, which extracts, stores, and analyzes key behavioral information to enrich character details. This method does not require modifications to the internal structure of LLMs and is compatible with various open-source and proprietary models. We conducted experiments comparing our approach with some of the best-performing long-story models currently available, and the results demonstrate that our model effectively improves the quality of long stories.