Abstract: Chinese relation extraction task aims to extract the semantic relation between entity pairs in Chinese text. The pre-trained language model(PLL) based method has achieved best results nowadays. As an alternative for lightweight computing, we propose a language model enhanced Chinese relation extraction method, which learns the language knowledge in the pre-trained language model based on multi-task learning. This paper conducts experiments on three Chinese relation extraction datasets and the results that, using only 1% of the parameters of the pre-trained language model, the proposed method can achieve 95% of PLL performance.