将下面改写的学术、精炼一点We processed the raw data as follows：firstly we exclude samples of ST ST and PT listed firms; secondly financial firms are excluded because of its special financial statements; thirdly sa

We adopted several steps to process the raw data. Firstly, we excluded samples of ST, *ST, and PT listed firms. Secondly, we excluded financial firms due to their unique financial statements. Thirdly, we excluded samples with missing key variables. Fourthly, we applied bilateral 1% winsorization to continuous variables to eliminate the impact of extreme values. Fifthly, we matched human-AI cooperation based on the city where the listed companies are located, while excluding samples from Tibet, Xinjiang, and certain areas of Yunnan. These steps resulted in a final sample size of 13,203 valid research samples.