metadata
license: mit
task_categories:
- image-to-text
- text-to-image
language:
- en
pretty_name: simons ARC (abstraction & reasoning corpus) solve half version 6
size_categories:
- 1K<n<10K
configs:
- config_name: default
data_files:
- split: train
path: data.jsonl
Version 1
ARC-AGI Tasks where the job is to identify where the top/bottom/left/right half of the object is located.
example count: 4-5.
test count: 1-2.
image size: 4-5.
Version 2
image size: 4-7.
Version 3
Earlier predictions added to some of the rows.
Version 4
Added fields: arc_task, test_index, earlier_output.
Version 5
Replaced RLE compressed response with raw pixel response. Argh I forgot to enable this. Using RLE compression.
image size: 4-15.
Version 6
Replaced RLE compressed response with raw pixel response.