A data scientist is experimenting with different hyperparameters for a machine learning model using MLflow. They want to compare the performance of different runs and select the best hyperparameters. What MLflow function should they use for hyperparameter tuning?