Thanks! Are the scores in some way linear here? As in, if model A is rated at 25 and model B at 50, does that mean I will have half the mistakes with model B? Get answers that are 2x more accurate? Or is it subjective?