Replication Data for: "Auditing the Hallucination Gap in Large Language Models for Utility Infrastructure Management"

This deposit contains the replication materials for the aforementioned study. Due to the sensitive nature of the original data, which details critical public water infrastructure, a synthetic dataset has been provided. This dataset was generated to mirror the statistical properties and relationships of the original data from Sugar Land, Texas, but contains entirely artificial records that cannot be traced back to the actual city infrastructure. Included are: 1. synthetic_water_asset_data.csv: A synthetic dataset replicating the structure of the original asset inventory. 2. synthetic_break_records.csv: A synthetic dataset replicating the structure of the original break records. 3. analysis_code.ipynb: The complete Python code used for data preprocessing, model training, evaluation, and text analysis. 4. README.txt: A detailed guide for running the code and interpreting the outputs.