Submissions are evaluated using Root Mean Squared Error (RMSE) between predicted and actual ratings.

A lower RMSE indicates a better model performance.