Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -5,8 +5,22 @@ task_categories:
|
|
| 5 |
- image-to-image
|
| 6 |
size_categories:
|
| 7 |
- 1K<n<10K
|
|
|
|
|
|
|
| 8 |
---
|
| 9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
A bad prediction, can be a hint towards solving the puzzle.
|
| 11 |
|
| 12 |
A bad prediction, can also be misleading.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
- image-to-image
|
| 6 |
size_categories:
|
| 7 |
- 1K<n<10K
|
| 8 |
+
language:
|
| 9 |
+
- en
|
| 10 |
---
|
| 11 |
|
| 12 |
+
# ARC Bad Predictions
|
| 13 |
+
|
| 14 |
+
Hypothesis: Is a stepwise refinement algorithm possible, that can improve on an earlier prediction?
|
| 15 |
+
|
| 16 |
+
This repo is intended for stepwise refinement algorithms. This dataset contains incorrect predictions that are somewhat close to the target.
|
| 17 |
+
I have manually inspected these predictions and removed the worst predictions. However there may still be bad predictions are are misleading.
|
| 18 |
+
|
| 19 |
A bad prediction, can be a hint towards solving the puzzle.
|
| 20 |
|
| 21 |
A bad prediction, can also be misleading.
|
| 22 |
+
|
| 23 |
+
The color values 0-9 are the same as the ARC-AGI colors. Color values outside this range sometimes occur, when a solver failed to predict that pixel.
|
| 24 |
+
|
| 25 |
+
The `dataset` and `task` correspond to a file found in the [arc-dataset-collection](https://github.com/neoneye/arc-dataset-collection) repo.
|
| 26 |
+
A few puzzles have multiple versions, these are suffixed with `_v2` and `_v3`.
|