Examining the Robustness of Homogeneity Bias to Hyperparameter Adjustments in GPT-4