ZICQ
中 Log in / Sign up
Newsroom LLMs & Foundation Models #BIM #LLMs & Foundation Models #AI Repair #Domain-Specific Models #Building Information Modeling

Intelligent BIM Design Defect Identification and Repair Framework Using Domain-Specific LLMs Released

Avatar of Mr.Xu

By Mr.Xu Compiled & Reviewed by Editorial

Published: · 4 views

中文阅读 (Chinese) English Version
Co-reading Select any sentence in the article to highlight, annotate, or copy a quote card. Join discussion ↓

Summary:This study introduces an integrated framework for identifying and repairing design defects in BIM using domain-specific large language models (LLMs). The framework leverages a BIM-to-Text conversion method, prompt learning with rule injection, few-shot prompting, and retrieval-augmented generation (RAG) to achieve high accuracy in defect identification and generate reliable repair suggestions. Experimental results demonstrate an 85% identification accuracy, outperforming traditional rule-based m


Background and Challenges

In the field of Building Information Modeling (BIM), existing methods for identifying and repairing design defects lack generality and struggle to handle the diversity of defect types. To address this, the study proposes an intelligent framework based on domain-specific large language models (LLMs) to achieve efficient identification and repair of BIM design defects.

Technical Approach

  1. BIM-to-Text Conversion: Utilizes a component-balanced chunking technique to convert BIM data into text format for LLM processing.
  2. Prompt Learning and Rule Injection: Enhances the model's defect identification capabilities through rule injection, few-shot prompting, and retrieval-augmented generation (RAG).
  3. Hallucination Control Strategy: Combines key identifier validation and token-length thresholds to reduce hallucinations and improve the reliability of generated suggestions.

Experimental Results

  • Defect Identification Accuracy: Achieves 85%, significantly outperforming the 70% accuracy of traditional rule-based methods.
  • Reasonable Repair Suggestion Generation Rate: Reaches 94%.
  • Hallucination Control Effectiveness: Boosts accuracy from 64% to 85% and eliminates 92.5% of hallucinations in a single intervention round.

Innovations and Advantages

  • High Generality: Capable of handling diverse defect types in BIM design.
  • High Efficiency: Demonstrates significant improvements in identification accuracy and repair suggestion generation rate compared to traditional methods.
  • High Reliability: Effectively reduces the generation of erroneous suggestions through the hallucination control strategy.

Industry Impact and Developer Recommendations

This framework provides a new technical path for the intelligent development of BIM, particularly in scenarios requiring efficient processing of complex design tasks. For developers, the following recommendations are advised:

  • Data Preprocessing: Ensure the accuracy and completeness of BIM data to improve conversion effectiveness.
  • Model Fine-Tuning: Fine-tune LLMs according to specific application scenarios to enhance identification and repair accuracy.
  • Hallucination Control: Continuously optimize the hallucination control strategy to minimize the generation of erroneous suggestions.

Future Outlook

In the future, this framework is expected to be widely applied in the Architecture, Engineering, and Construction (AEC) industry, driving the intelligent development of BIM design. As LLM technology continues to advance, the performance and reliability of the framework will also be further enhanced.


Source: ArXiv NLP/LLM (cs.CL) (2026-09-01)

— END —

Tags: #BIM #LLMs & Foundation Models #AI Repair #Domain-Specific Models #Building Information Modeling

Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.

What's your take on this story?
🚀 Explore & Submit Works

Community Comments

Loading live comments and annotations…