CrossTest, a new large-scale challenging dataset for CLRC (Cross-Lingual Reading Comprehension). To our knowledge, it is the first public dataset which aims to test machines on their CLRC ability, especially the ability to answer the question written in another language without translation. CrossTest is a close-style dataset which requires the reader to fill in a missing word (we provide ten noun an- swer candidates for each sample) in a sentence written in target language by reading a given passage written in source language. It consists of two dual sub-datasets: QCPE Test and QEPC Test. Questions are written in Chinese or English for QCPE Test and for QEPC Test respectively.