Existing compression artifacts reduction methods aim to restore images on pixel-level, which can improves human visual experience. However, in many applications, large-scale images are collected not for visual examination by human. Instead, they are used for many high-level vision tasks usually by Deep Neural Networks (DNN). One fundamental problem here is whether existing artifacts reduction methods can help DNNs improve the performance of the high-level tasks. In this paper, we find that these methods have limited performance improvements to high-level tasks, even bring negative effects. Therefore, inspired by the teacher-student network framework, we propose a compression artifacts reduction framework (ARF) for DNN. In addition, we generalize our method to the unsupervised setting (U-ARF) where the correspond original images are unavailable in training. Extensive experiments on image classification and person re-identification indicate the proposed methods can help DNNs improve performance on the highly compressed images significantly.