ARC-E

ARC-E (ARC Easy) is a baseline reasoning benchmark consisting of straightforward multiple-choice science questions that test fundamental knowledge and basic reasoning skills. Designed to ensure AI systems have the essential understanding needed for scientific reasoning, it serves as an entry-level evaluation before progressing to more challenging assessments. The benchmark focuses on simple inference capabilities and basic scientific comprehension that most competent AI systems should master.