Datasets:
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
## Summary
|
| 2 |
+
This dataset contains spectral classification tasks built for tool-using models.
|
| 3 |
+
Each example includes a full-spectrum image, a chat-style prompt, and ground-truth
|
| 4 |
+
labels for binary decisions. The metadata also provides the underlying spectral
|
| 5 |
+
arrays (wavelength and flux) to support visualization or custom feature extraction.
|
| 6 |
+
|
| 7 |
+
## Data Files
|
| 8 |
+
All files are under `benchmark/data` and share the same schema.
|
| 9 |
+
File suffixes correspond to target classes:
|
| 10 |
+
|
| 11 |
+
- `cc` -> Carbon Star (C-type) (data_source: `carbon`)
|
| 12 |
+
- `cv` -> Cataclysmic Variable (CV)
|
| 13 |
+
- `gm` -> M-type Giant (gM)
|
| 14 |
+
- `ss` -> S-type Star
|
| 15 |
+
- `wd` -> White Dwarf (WD)
|
| 16 |
+
|
| 17 |
+
## Splits and Sizes
|
| 18 |
+
|
| 19 |
+
| split | file | rows |
|
| 20 |
+
| --- | --- | --- |
|
| 21 |
+
| train | train_cc.parquet | 671 |
|
| 22 |
+
| train | train_cv.parquet | 711 |
|
| 23 |
+
| train | train_gm.parquet | 658 |
|
| 24 |
+
| train | train_ss.parquet | 706 |
|
| 25 |
+
| train | train_wd.parquet | 618 |
|
| 26 |
+
| test | test_cc.parquet | 1321 |
|
| 27 |
+
| test | test_cv.parquet | 1458 |
|
| 28 |
+
| test | test_gm.parquet | 1321 |
|
| 29 |
+
| test | test_ss.parquet | 1442 |
|
| 30 |
+
| test | test_wd.parquet | 1212 |
|
| 31 |
+
|
| 32 |
+
Totals: train 3364, test 6754.
|
| 33 |
+
|
| 34 |
+
## Data Structure
|
| 35 |
+
Each example contains the following fields:
|
| 36 |
+
|
| 37 |
+
- `images`: list with a single image object
|
| 38 |
+
- `data_source`: class key string, e.g. `carbon`, `cv`, `gm`, `ss`, `wd`
|
| 39 |
+
- `prompt`: list of chat messages
|
| 40 |
+
- `role`: message role (`system`, `user`, etc.)
|
| 41 |
+
- `content`: message text (the user message includes an `<image>` placeholder)
|
| 42 |
+
- `reward_model`: struct with ground-truth labels
|
| 43 |
+
- `ground_truth`: string label (typically `YES` or `NO`)
|
| 44 |
+
- `type`: label source (e.g. `rule`)
|
| 45 |
+
- `agent_name`: string identifier for the agent configuration
|
| 46 |
+
- `extra_info`: metadata
|
| 47 |
+
- `index`: unique id string
|
| 48 |
+
- `name`: spectrum file name
|
| 49 |
+
- `tools_kwargs.spectral_visualization_tool.create_kwargs`:
|
| 50 |
+
- `flux`: list of floats (length ~3909)
|
| 51 |
+
- `wavelength`: list of floats (length ~3909)
|
| 52 |
+
- `redshift`: float
|
| 53 |
+
|
| 54 |
+
## Notes
|
| 55 |
+
- `images` contains the full-spectrum overview image for the example.
|
| 56 |
+
- `extra_info` provides raw arrays to recreate spectrum plots.
|