This article analyzes the psychometric foundations of the IRT methodology for identifying cognitive misconceptions of 7th-9th graders. Based on Uzbekistan’s PISA-2022 and TIMSS-2023 results, low performance in the “reasoning” cognitive domain is linked to misconceptions. A comparative analysis of IRT models (1PL, 2PL, 3PL), possibilities of cognitive diagnostic assessment (CDA) and the Q-matrix, PSM methods, and cognitive intervention strategies (“cognitive conflict”) are presented. The analysis serves as a methodological basis for identifying systemic gaps and developing intervention programs.