neoneye's picture
Update README.md
c1d540a verified
|
raw
history blame
1.09 kB
---
license: mit
task_categories:
- image-to-text
- text-to-image
language:
- en
pretty_name: simons ARC (abstraction & reasoning corpus) solve rotate version 10
size_categories:
- 1K<n<10K
configs:
- config_name: default
data_files:
- split: train
path: data.jsonl
---
# Version 1
ARC-AGI Tasks where the image gets rotated cw/ccw/180 and transposed.
The image sizes are between 1 and 4 pixels.
Predict the number of rows in the output image.
# Version 2
image size: 1-5.
# Version 3
image size: 1-5.
Added `flipx` and `flipy` transformations.
# Version 4
image size: 1-5.
number of tests: 1-2. Previously there were always just 1 test.
Added `flipa` and `flipb` transformations, that flips over the diagonal.
# Version 5
image size: 1-5.
number of tests: 1-2.
# Version 6
image size: 1-13.
# Version 7
Earlier predictions added to some of the rows.
# Version 8
Earlier predictions with focus on repair 1 bad pixel.
# Version 9
Added fields: `arc_task`, `test_index`, `earlier_output`.
# Version 10
Replaced RLE compressed response with raw pixel response.