b) In Forward Training, we perform multiple policy updates along each trajectory, thus enabling the model to update its policy very quickly