Revisiting Fairness in Multitask Learning: A Performance-Driven Approach for Variance Reduction | Read Paper on Bytez