I want to add noise to the reward distribution I have. In what format should the reward distribution be represented for VW to understand and what methods are available in VW to
