1 paper touches this idea.
Papers
An AI that teaches itself gets sharper but stops exploring
Related concepts
Ideas that show up alongside Out-Of-Distribution Generalization.