Optimized code generation in BIM with retrieval-augmented LLMs

Thumbnail Image

Files

master_Naveed_Jahanzaib_2025.pdf (1.97 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Science | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-02-27

Department

Major/Subject

Computer Science

Mcode

Degree programme

Master's Programme in Computer, Communication and Information Sciences

Language

en

Pages

64

Series

Abstract

The growing interest in Large Language Models (LLMs), such as Generative Pre-Trained Transformers (GPT), has spanned various industries, including software development, construction, and medicine. In the construction sector, LLMs hold potential for enhancing Building Information Modeling (BIM) software, which is widely used to create digital representations of physical structures for planning and design purposes. This study explores how LLMs can improve BIM by enabling users to implement architectural features through text-based prompts. To achieve this, LLMs must query relevant knowledge bases to generate macros (code files) that modify BIM templates. Given that LLMs are trained on generalpurpose data, Retrieval-Augmented Generation (RAG) techniques can be employed to enhance their ability to access domain-specific information. While LLMs show promise, further research is required to determine whether GPT-generated macros can be effectively integrated into BIM workflows to improve productivity. Additionally, there is a shortage of studies assessing the reliability of RAG in generating accurate code for BIM applications. This research aims to address these gaps by exploring various RAG implementation methods within GPT models and benchmarking the performance of the resulting macros. The study begins with a review of the relevant background on LLMs and RAG, followed by the research methodology, analysis of results, and conclusions. Ultimately, this work lays the foundation for leveraging LLMs in BIM software to enhance efficiency and accuracy.

Description

Supervisor

Jung, Alex

Thesis advisor

Haikola, Ville

Keywords

LLM, RAG, BIM, large learning models, trimble, AI

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202505203933

Collections

[dipl] Perustieteiden korkeakoulu / SCI