Abstraction and Reasoning Corpus (ARC)

The Abstraction and Reasoning Corpus (ARC) is a benchmark that contains tasks requiring solvers to infer abstract patterns from limited examples, used to evaluate general AI capabilities and reasoning skills.

Papers using Abstraction and Reasoning Corpus (ARC) (3)

Your Reasoning Benchmark May Not Test Reasoning: Revealing Perception Bottleneck In Abstract Reasoning Benchmarks2025ARC Is A Vision Problem!2025GIFARC: Synthetic Dataset For Leveraging Human-intuitive Analogies To Elevate AI Reasoning2025