license: mit
task_categories:
- image-to-text
- text-to-image
language:
- en
pretty_name: simons ARC (abstraction & reasoning corpus) solve color version 11
size_categories:
- 10K<n<100K
configs:
- config_name: default
data_files:
- split: train
path: data.jsonl
Version 1
ARC-AGI Tasks where the colors gets manipulated.
Currently it's two-color images, where the transformation is to swap colors.
The image sizes are between 1 and 5 pixels.
Predict the number of rows in the output image.
Version 2
Number of test: 1-2. Previously it was always 1 test.
Version 3
input image size: 1-3.
Number of tests: 1.
Identify most popular color, and least popular color. The output size is always 1x1.
Version 4
input image size: 1-4.
Number of tests: 1.
Identify most popular color, and least popular color. The output size is always 1x1.
Version 5
input image size: 1-5.
Number of tests: 1-2.
Identify most popular color, and least popular color. The output size is always 1x1.
Version 6
input image size: 1-5.
Number of tests: 1-2.
Identify most popular color, and least popular color. Multiple output sizes: output size is 1x1, and same output size as input size.
Swap colors.
Version 7
Focus on generate_task_replace_color.
image size: 3-6.
padding size: 1-5.
Version 8
Focus on generate_task_replace_color.
image size: 3-8.
padding size: 1-10.
Version 9
Focus on generate_task_replace_color.
image size: 3-10.
padding size: 1-20.
Version 10
Enabled all the task generators.
Version 11
Focus on generate_task_replace_color_pairs_with_different_palettes.
image size: 3-5.
padding size: 1-4.