Abstract: The validity of Gaokao(Chinese college entrance examination), a selective test of academic evaluation, depends on its identification of the variability of students' transferable ability in problem-solving. However, the raw score in academic evaluation does not reflect the actual level of students' ability, which is a latent variable. In order to make the results of academic evaluation more valid, this study constructs a moderated model based on latent variable by treating a student with full score in ability as a reference. Raw score of a particular question is re-weighed according to the difficulty of the question. The moderated model based on the latent variable was applied to the data analysis of an 11-school-league examination, with a total of 9,008 high school students participating in 10 subjects tests. The results show that:a) the adjusted score is more normal than the raw score; b) the ability score is more stable than the raw score; c) the total score has a high correlation with the ability score; d) individually, there is a big difference between the raw score and the ability score.

Key words: academic assessment, reform of Gaokao, validity, transfer, ability, latent variable