Abstract: With the development of the internet and online education, limitations of time and space were overcome, allowing people to learn anytime and anywhere. A vast amount of learning records was generated through various online learning and competition platforms. By effectively utilizing these records, computerized adaptive testing(CAT)enabled the customization of personalized assessments for examinees, achieving the goal of "tailored instruction". This paper aimed to comprehensively review the current development and state-of-the-art work in CAT, provide an outlook on future directions, and help future researchers and practitioners gain a systematic understanding of CAT. First, the background and theoretical foundations of CAT were introduced, followed by a formal description of CAT. Then, from a technical perspective, existing CAT methods were categorized into two types: CAT methods for selection process and CAT methods for ability estimation. A detailed overview of these two types of CAT methods was provided. Next, commonly used public datasets and evaluation metrics in CAT were compiled, with each dataset's source and relevant information described. Finally, the future research directions of CAT were discussed, and conclusions were drawn.

Key words: computerized adaptive testing, item response theory, question selection algorithm, ability estimation method, educational data minining