Prompt Sensitivity (Performance Variance) is a crucial evaluation metric for prompt engineering tasks in the fields of large language models and inference. It measures model robustness to prompts by quantifying performance fluctuations caused by different prompt phrasings. Lower variance is better, indicating that the model is less sensitive to changes in prompts. This metric is typically reported as a dimensionless score. It is widely used in relevant benchmarks and papers, effectively reflecting model performance in this dimension. Lower values are better.