ARC-AGI

ARC-AGI (Abstraction and Reasoning Corpus for Artificial General Intelligence) is the foundational benchmark measuring progress towards general intelligence through fluid reasoning tasks. It evaluates AI systems' ability to solve novel visual puzzles requiring abstract pattern recognition, spatial reasoning, and logical inference. The benchmark tests core cognitive abilities like few-shot learning and generalization without relying on prior knowledge, serving as a critical measure of human-like reasoning capabilities.