I. Introduction
The XAI evaluation framework has been paid more and more attention under the condition that the XAI methods have been widely accepted. At present, the interpretability of the model is obviously focused on the field of ML and DL, especially when their models are used in medical, financial, military and other fields that require high transparency and trust, the interpretability of the model is even necessary. However, black-box models such as neural networks [1] and ensemble models [2] are not inherently interpretable. Therefore, the XAI method is absolutely needed to promote the more efficient application of black-box models with higher accuracy in real life. The XAI evaluation framework can select XAI methods that are more suitable for the model, which further improves the credibility of the black-box model.