“Adaptive learning” as a mechanistic candidate for reaching optimal task-set representations flexibly | Publicación