Skip to content

ARC-AGI-1 what it measures, how it is scored and who leads it

Reasoning · ARC Prize Foundation (created by François Chollet) · introduced Nov 5, 2019 · Tasks solved

About this benchmark

Task
Infer the rule behind a few input-output grid pairs and apply it to a new grid.
Dataset
Public training and evaluation sets of abstract grid puzzles, plus semi-private and private evaluation sets.
Method
Two attempts per task (pass@2); an answer counts only if every cell of the output grid is right.
Metric
Tasks solved
Organization
ARC Prize Foundation (created by François Chollet)
Introduced
Nov 5, 2019
Official leaderboard
arcprize.org/leaderboard

Versions

Newest first. New versions are added, never rewritten.

VersionDate
Public leaderboardARC Prize opens a public leaderboard with semi-private evaluation.Jun 27, 20242 years ago
1Published with the paper as the Abstraction and Reasoning Corpus.Nov 5, 20196 years ago

Sources: each benchmark's paper (arXiv) and its official site or leaderboard. Scores are the boards captured here every day, exactly as published.

New #1s by email

Saturdays, only in weeks when a leaderboard has a new #1.

Double opt-in. Unsubscribe any time.