In this presentation, we will explore the multiple testing of the general regression framework, which aims to investigate the correlation between a univariate response and a p-dimensional predictor. We will discuss recent advancements in using the simple data perturbation method of data splitting to control the false discovery rate (FDR) in fitting quantile regression and index models. The data splitting procedure divides the data randomly into two halves and computes a statistic reflecting the consistency of the parameter estimates (such as regression coefficients) between the two sets. By utilizing statistics that are symmetrically distributed around zero for null features, FDR control can be achieved. Notably, the construction of mirror statistics can circumvent challenges caused by variance estimation and has demonstrated numerical advantages over existing methods. This is a Joint work with Jun Liu, Xin Xing, Yan Yu and Tianhai Zu.