Abstract
As large language models (LLMs) become increasingly integrated into cultural heritage research, practices such as artifact reconstruction, content generation, and virtual restoration are rapidly advancing. Generative artificial intelligence is unlocking new narrative potential for jiandu (bamboo and wooden slips), offering innovative approaches to cultural expression and historical reconstruction. However, general-purpose language models often face limitations in domain knowledge and semantic reliability, making it difficult to ensure accuracy and traceability in jiandu-based generation. To address these challenges, this study proposes a virtual restoration framework for jiandu narratives based on Retrieval-Augmented Generation (RAG) and agent mechanisms. Using the Liye Qin slips as a case study, the LIYE-Agent system integrates structured knowledge, semantic retrieval, and task-oriented generation to achieve coherent scene reconstruction. Findings from discipline-informed participant evaluation demonstrate its effectiveness in enhancing generative accuracy and consistency, providing methodological and technical support for generative narrative design and the digital dissemination of cultural heritage.
Keywords
Retrieval-Augmented Generation(RAG); Agent-based System; Liye Qin slips; Generative Narrative
DOI
https://doi.org/10.21606/drs.2026.1738
Citation
Tian, Q., Yao, Y., Chen, Z., and Zhang, H. (2026) Designing Narrative Reconstruction from Bamboo Slips: A RAG-Agent Approach for Cultural Scene Generation, in Simeone, L., Gray, C. M., Verhoeven, A., de Götzen, A., Bakırlıoğlu, Y., Zohar, H., Stead, M., and Buwert, P. (eds.), DRS2026: Edinburgh, 8–12 June, Edinburgh, United Kingdom. https://doi.org/10.21606/drs.2026.1738
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Included in
Designing Narrative Reconstruction from Bamboo Slips: A RAG-Agent Approach for Cultural Scene Generation
As large language models (LLMs) become increasingly integrated into cultural heritage research, practices such as artifact reconstruction, content generation, and virtual restoration are rapidly advancing. Generative artificial intelligence is unlocking new narrative potential for jiandu (bamboo and wooden slips), offering innovative approaches to cultural expression and historical reconstruction. However, general-purpose language models often face limitations in domain knowledge and semantic reliability, making it difficult to ensure accuracy and traceability in jiandu-based generation. To address these challenges, this study proposes a virtual restoration framework for jiandu narratives based on Retrieval-Augmented Generation (RAG) and agent mechanisms. Using the Liye Qin slips as a case study, the LIYE-Agent system integrates structured knowledge, semantic retrieval, and task-oriented generation to achieve coherent scene reconstruction. Findings from discipline-informed participant evaluation demonstrate its effectiveness in enhancing generative accuracy and consistency, providing methodological and technical support for generative narrative design and the digital dissemination of cultural heritage.