README.md

metadata

license: mit
task_categories:
  - image-to-text
  - text-to-image
language:
  - en
pretty_name: simons ARC (abstraction & reasoning corpus) solve scale version 9
size_categories:
  - 1K<n<10K
configs:
  - config_name: default
    data_files:
      - split: train
        path: data.jsonl

Version 1

ARC-AGI Tasks where the images gets scaled up/down in both x and y direction.

example count: 2-4.

test count: 1-2.

image size: 3-10.

scale factor: 1-3.

Version 2

image size: 1-20.

scale factor: 1-7.

Version 3

image size: 1-30.

scale factor: 1-7.

Version 4

Added a few noise to the images.

image size: 1-10.

scale factor: 1-7.

Only scale down.

Number of noise pixels per pixel cell: 0-2.

Version 5

More noisy images for down scaling.

image size: 1-12.

Number of noise pixels per pixel cell: 0-half.

Version 6

Earlier predictions added to some of the rows.

Version 7

Added fields: arc_task, test_index, earlier_output.

Version 8

Replaced RLE compressed response with raw pixel response.

image size: 1-5.

scale factor: 1-7.

Version 9

image size: 1-7.

scale factor: 1-3.