# ARC-AGI

> ARC-AGI (the Abstraction and Reasoning Corpus) is a benchmark of visual grid puzzles in which a model must infer a transformation rule from a few examples and apply it to a new grid, designed to measure how efficiently a system acquires new skills.

- Identifier: PTL-0192
- Category: Benchmarks
- Canonical URL: https://protologue.com/t/arc-agi/
- Also known as: ARC, Abstraction and Reasoning Corpus, ARC-AGI-2, ARC Prize
- Introduced: 2019

## Description

The puzzles are easy for most people but require reasoning that cannot be memorized, because each task is novel. François Chollet introduced it in 2019 alongside a definition of intelligence as skill-acquisition efficiency. Successive versions, such as ARC-AGI-2, were made harder as systems improved, and many results are reported with explicit compute costs.

## Broader terms

- [Benchmark](https://protologue.com/t/benchmark/)

## Related terms

- [Test-Time Compute Scaling](https://protologue.com/t/test-time-compute-scaling/)
- [Reasoning Model](https://protologue.com/t/reasoning-model/)

## Sources

- Chollet (2019). On the Measure of Intelligence. https://arxiv.org/abs/1911.01547
- Chollet et al. (2025). ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems. https://arxiv.org/abs/2505.11831

## Cite this entry

Protologue. (2026). ARC-AGI. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0192). https://protologue.com/t/arc-agi/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)