ARC-AGI (Abstraction and Reasoning Corpus for Artificial General Intelligence) is a benchmark created by François Chollet designed to measure fluid intelligence in AI systems—the ability to efficiently acquire new skills from limited examples, testing whether AI can truly generalize.
Created the most challenging and theoretically grounded benchmark for measuring AI general intelligence; resisted AI system performance for years before GPT-o1 made progress on it.
No conceptual posterity found.
Discussing
Ask anything about ARC — their ideas, influences, contradictions, legacy.
Sign in to start a conversation.