Is it possible for Ray Tune to produce hyperparameters worse than default ones?

I have tried using Ray tune for my YOLOv8 models, and so far, all the “tuned” models are performing worse than models trained on default hyperparameters. I am selecting the hyperparameters of the tuning trial that gave the highest ‘mAP50 score’ ; it seems the only other available metrics are "precision(B)', ‘recall(B)’, and ‘mAP50-95(B)’. What could be the reason for this? I have invested a lot of time into getting it working in my model development workflow, and so far it seems to have served the opposite effect :frowning: