Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,75 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
# Dataset Card for "JDWebProgrammer/arg-agi-augmented"
|
| 6 |
+
|
| 7 |
+
## Dataset Description
|
| 8 |
+
|
| 9 |
+
### Overview
|
| 10 |
+
This dataset is an augmented version of grids extracted from the [ARC-AGI dataset](https://huggingface.co/datasets/dataartist/arc-agi) (Abstraction and Reasoning Corpus). It focuses on **individual grids** rather than full tasks or games, providing an expanded collection for pretraining and testing models like autoencoders (AEs) or latent-space reasoners.
|
| 11 |
+
|
| 12 |
+
- **Source**: Derived from the `training` split of ARC-AGI (all demonstration and test grids).
|
| 13 |
+
- **Augmentations**: Each original grid is expanded with 5 transformations (horizontal flip, vertical flip, 90°/180°/270° rotations), resulting in 6 variants per grid (original + 5 augments).
|
| 14 |
+
- **Key Note**: This is **not the full games/tasks** from ARC-AGI. It contains only the raw, augmented grids (as 2D lists of integers 0-10) for standalone use in perceptual pretraining or reconstruction testing. Use the original ARC-AGI for full few-shot reasoning tasks.
|
| 15 |
+
|
| 16 |
+
### Dataset Structure
|
| 17 |
+
- **Format**: Hugging Face `Dataset` object.
|
| 18 |
+
- **Splits**: Single split (`train`) with one field:
|
| 19 |
+
- `augmented_grids`: List of 2D lists (grids). Each grid is `[[int, ...], ...]` (H x W, values 0-10).
|
| 20 |
+
- **Size**: ~48,000 grids (from ~400 ARC training tasks × ~4 grids/task × 6 augments).
|
| 21 |
+
- **Metadata**: See `metadata.json` for stats (original grids, augmentation factor).
|
| 22 |
+
|
| 23 |
+
Example grid entry:
|
| 24 |
+
```python
|
| 25 |
+
augmented_grids[0] = [[0, 1, 0], [1, 0, 1], [0, 1, 0]] # Example 3x3 grid
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
### Usage
|
| 29 |
+
Load and use for AE pretraining:
|
| 30 |
+
```python
|
| 31 |
+
from datasets import load_dataset
|
| 32 |
+
ds = load_dataset("JDWebProgrammer/arc-agi-augmented")
|
| 33 |
+
grids = ds['train']['augmented_grids'] # List of all grids
|
| 34 |
+
|
| 35 |
+
# Example: Batch grids for AE
|
| 36 |
+
def grid_to_tensor(grid):
|
| 37 |
+
h, w = len(grid), len(grid[0])
|
| 38 |
+
return torch.tensor(grid, dtype=torch.float).view(1, -1) / 10.0 # Normalize 0-1
|
| 39 |
+
|
| 40 |
+
batch = torch.cat([grid_to_tensor(g) for g in grids[:32]]) # Batch of 32
|
| 41 |
+
# Feed to AE: z = ae.encode(batch); recon = ae.decode(z)
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
Ideal for:
|
| 45 |
+
- Pretraining perceptual models.
|
| 46 |
+
- Testing reconstruction accuracy (compare original vs. augmented).
|
| 47 |
+
- Data augmentation for fluid intelligence tasks (e.g., ARC-like pattern inference).
|
| 48 |
+
|
| 49 |
+
### Generation
|
| 50 |
+
- Extracted all input/output grids from ARC-AGI `training` split demos/tests.
|
| 51 |
+
- Applied deterministic augmentations (flips/rotations) to expand variety without labels.
|
| 52 |
+
- No synthetic generation — pure augmentation of real ARC data.
|
| 53 |
+
|
| 54 |
+
### Limitations
|
| 55 |
+
- Grids only (no task structure/context) — not for end-to-end ARC solving.
|
| 56 |
+
- Augmentations preserve structure but may introduce artifacts (e.g., rotations on asymmetric grids).
|
| 57 |
+
- Values 0-10 (ARC standard); normalize for models.
|
| 58 |
+
|
| 59 |
+
### License
|
| 60 |
+
- Based on ARC-AGI (CC BY-SA 4.0) — inherits same license.
|
| 61 |
+
- Augmentations: MIT (free for research/commercial).
|
| 62 |
+
|
| 63 |
+
### Citation
|
| 64 |
+
```bibtex
|
| 65 |
+
@misc{dataartist/arc-agi,
|
| 66 |
+
title = {Augmented ARC-AGI Grids for Pretraining},
|
| 67 |
+
author = {dataartist},
|
| 68 |
+
year = {2025},
|
| 69 |
+
url = {https://huggingface.co/datasets/your_username/arc-augmented-grids}
|
| 70 |
+
}
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
---
|
| 74 |
+
|
| 75 |
+
*Generated for pretraining perceptual models on ARC-style puzzles. Not a substitute for full ARC tasks.*
|