lalalune / arc-datasets

A collection of datasets for ARC AGI Challenge

Multiple datasets for ARC (Abstraction and Reasoning Corpus)

Two of the datasets were generated with data from https://github.com/lalalune/arc-wfc-data and https://github.com/lalalune/rulepredictor

Some of the datasets were pulled from here: https://github.com/neoneye/arc-dataset-collection

BabyARC was pulled form here: https://github.com/frankaging/BabyARC

Data structure

All data is stored in datasets//data/training and datasets//data/evaluation (if avalailable). Only a few have evaluation data.

Each dataset have its own license, see the readme/license for details.