# fchollet/ARC-AGI

The Abstraction and Reasoning Corpus

Repository: https://github.com/fchollet/ARC-AGI
Canonical: https://ross.abutalabs.com/products/arc-agi
Language: JavaScript
License: Apache-2.0
License Family: permissive
Topics: artificial-intelligence, program-synthesis, psychometrics, intelligence-testing
Last push: 2025-04-04T21:28:40+00:00

## Health v2 (maintenance only)
Score: 29/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 14, release rhythm 8, longevity 100
- inputs: {"age_days": 2494, "days_push": 516, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4813, forks 725 (observed 2026-08-28T04:09:00.299718+00:00)

## What it is
The Abstraction and Reasoning Corpus (ARC-AGI-1), a benchmark dataset of 800 JSON grid-based tasks designed to test human-like fluid general intelligence and program synthesis ability. It includes a browser-based interface for humans to attempt tasks manually.

## Use cases
- benchmark an AI system on abstract reasoning tasks
- evaluate program synthesis algorithms
- train models on ARC-relevant cognitive priors
- let humans try solving intelligence test puzzles in a browser
- compare human and AI general fluid intelligence

## When to avoid
- you need large-scale training data for supervised learning
- you want a conventional NLP or vision benchmark
- you need production-ready ML tooling rather than raw task data

## Facets
- artifact type: dataset
- maturity: stable
- function: machine-learning, data-science, benchmarking
- domain: artificial-intelligence, machine-learning
- platform: cross-platform, browser
- tags: benchmark, program-synthesis, psychometrics, agi, intelligence-testing, json-tasks, grid-puzzles, algorithms

## Member repositories
- fchollet/ARC-AGI (main) score 29

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:00.299718+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:18:35.113967+00:00, confidence not recorded.
  - readme: https://github.com/fchollet/ARC-AGI (fetched 2026-08-28T04:09:00.299718+00:00, sha 2ac5a657c6f2)
- Data as of 2026-08-30T08:39:29.467469+00:00.