| id |
caadria2025_40 |
| authors |
liu, deli, Zhou, Xiaoping and Li, Yu |
| year |
2025 |
| title |
Enhancing Natural Language Retrieval of BIM Data through Integration of Large Language Models with Multi-Agent Systems |
| source |
Dagmar Reinhardt, Christiane M. Herr, Anastasia Globa, Jielin Chen, Taro ?Narahara, Nicolas Rogeau (eds.), ARCHITECTURAL INFORMATICS - Proceedings of the 30th CAADRIA Conference, Tokyo, 22-29 March 2025, Volume 3, pp. 91–100 |
| summary |
Building Information Modeling (BIM) data plays a pivotal role in modern construction projects. However, retrieving relevant information from BIM can be challenging, often requiring significant technical expertise and tool proficiency. In this paper, we propose a novel approach to address the difficulties associated with BIM data retrieval by leveraging the capabilities of LLMs and multi-agent systems. We designed a multi-agent system to facilitate efficient BIM data retrieval, integrating code generation, checking, error reflection, and execution. This multi-agent system can generate code and execute it to retrieve BIM data based on users' queries. Additionally, it can provide natural language answers to users based on the retrieved BIM data. Three distinct LLMs were selected to build the multi-agent and were compared to determine the optimal one. We also prepared documentation of the IfcOpenShell Application Programming Interface (API) to guide the code generation agent. This initiative aims to enhance the performance of the LLMs in professional tasks by ensuring clarity and precision in code generation. An evaluation method was developed to assess system performance. Lastly, a web-based IFC Model View was designed for enhanced user interaction and three-dimensional visualization of retrieved IFC models. |
| keywords |
Building Information Modeling (BIM), Large Language Models (LLMs), Multi-agent system, Data retrieve, Natural Language |
| series |
CAADRIA |
| full text |
 file.pdf (1,384,406 bytes) |
| references |
Content-type: text/plain
|
| last changed |
2025/03/21 12:08 |
|
