A Study on Large Language Model-Based Approach for Construction Contract Risk Detection

Construction projects typically involve large-scale operations and are subject to complex external conditions, making it essential to safeguard the interests of contractor enterprises through well-crafted contract clauses. However, the current reliance on expert judgment for identifying contract risks presents several challenges, including lengthy processing times, heavy workloads, and inconsistent results. To address these issues, this study introduces a Large Language Model (LLM)-based approach for automating the identification of risks in construction contracts. The proposed method was rigorously validated on 26 actual contracts, achieving an average accuracy of 76.7% across four state-of-the-art LLMs. This research advances the application of LLMs in construction contract management, providing practical solutions to existing challenges and setting the stage for further exploration in LLM-driven contract analysis.