Reward Learning From Preference With Ties