Abstract page for arXiv paper 2610.22379: When is a large language model a good calibration agent? Physics-labeled prompting for budgeted engineering simulator calibration