Тұрақты атрибуция сынып түсіндірмесі

AbstractInput saliency methods have recently become a popular tool for explaining predictions of deep learning models in NLP. Nevertheless, there has been little work investigating methods for aggregating prediction-level explanations …...