CRUX-O

CRUX-O is a code understanding benchmark that focuses on output prediction tasks for evaluating AI models' program comprehension abilities. This benchmark tests models' capacity to analyze code logic, trace execution paths, and accurately predict program outputs. CRUX-O serves as a fundamental evaluation tool for measuring code reasoning and execution simulation capabilities in language models.