Can we learn better with hard samples?

Can we learn better with hard samples?

Papers citing "Can we learn better with hard samples?"