Detection Is a Separate From the Learning Representation

In our method, since we introduce fairness notion as a integrated method, we typically out perform most of the competitors on the fairness task and so on. And there we also notice that the performance doesn't deviate much from the base model its or it deviates least from the base moral compared to any other competitor. So in that sense, what we show is that by following our deciderata and by incorporating the notions of fairness that we have introduced, we are able to achieve a very nice balance between fairness as well as performance.