In this study, the reasoning evaluation framework were divided into two task categories: Basic Logical Reasoning and Contextual Reasoning. Together, these categories captured a model’s overall performance, spanning from fundamental reasoning skills to more advanced reasoning abilities.