Can the model refuse to answer when it cannot guarantee the result？

[OmegaDING]

Thanks for your work， 可否在输出结果时加入判断，若置信度较低或不曾在训练集见过类似的问题就让模型拒绝回答问题？