If you use an ensemble of different base models, is it necessary to tune the hyper parameters of all base models to improve the ensemble performance?

• A. yes
• B. no
• C. can’t say