Structure of the Elementary Subtraction Skill within 20, Verified via Conversational AI Models (Data Array of a Pedagogical Experiment in 4th Grades, 1993–1995)

The publication describes the structure of the elementary subtraction skill within 20, verified using large conversational models (LLMs). The presented dataset reflects the state of fourth-graders' skills during the experiment period (1993–1995); however, the author's long-term observations indicate a fairly stable preservation of its structure under modern conditions. The work demonstrates the applied potential of neural network technologies in pedagogical measurements and statistical data processing.