Recent findings reveal that transformer attention mechanisms exhibit significant deficits in executive control. This impairment hinders the ability of models to perform complex reasoning tasks accurately.