Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
json
Languages:
English
Size:
10K - 100K
License:
Add dataset card
Browse files
README.md
ADDED
|
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
task_categories:
|
| 6 |
+
- text-classification
|
| 7 |
+
pretty_name: System One Mini Synthetic Diagnostics
|
| 8 |
+
configs:
|
| 9 |
+
- config_name: default
|
| 10 |
+
data_files:
|
| 11 |
+
- split: train
|
| 12 |
+
path: data/train.jsonl
|
| 13 |
+
- split: validation
|
| 14 |
+
path: data/validation.jsonl
|
| 15 |
+
- split: calibration
|
| 16 |
+
path: data/calibration.jsonl
|
| 17 |
+
- split: development_renderer
|
| 18 |
+
path: data/development_renderer.jsonl
|
| 19 |
+
- split: test
|
| 20 |
+
path: data/test.jsonl
|
| 21 |
+
---
|
| 22 |
+
|
| 23 |
+
# System One Mini Synthetic Diagnostics
|
| 24 |
+
|
| 25 |
+
Deterministic synthetic controlled-intervention summaries used by `DavidHatley/system-one-mini`. The records contain no real coding-agent traces, private repositories, personal data, external model outputs, or teacher labels.
|
| 26 |
+
|
| 27 |
+
**This dataset and model are independent research artifacts, not reproductions of Jev or RLCD.**
|
| 28 |
+
|
| 29 |
+
## Splits
|
| 30 |
+
|
| 31 |
+
| Split | Rows |
|
| 32 |
+
|---|---:|
|
| 33 |
+
| `train` | 20,000 |
|
| 34 |
+
| `validation` | 2,000 |
|
| 35 |
+
| `calibration` | 2,000 |
|
| 36 |
+
| `development_renderer` | 2,000 |
|
| 37 |
+
| `test` | 2,000 |
|
| 38 |
+
|
| 39 |
+
`development_renderer` is the former sealed renderer split that was subsequently used to compare candidates. It must not be described as an unbiased final test. `test` is a separate renderer written and frozen after model selection, then evaluated once for the initial release.
|
| 40 |
+
|
| 41 |
+
## Fields
|
| 42 |
+
|
| 43 |
+
- `id`: SHA-256 fingerprint of the latent synthetic situation
|
| 44 |
+
- `state`: rendered diagnostic summary
|
| 45 |
+
- `labels`: five numeric labels in schema order
|
| 46 |
+
- Five named string label fields
|
| 47 |
+
- `latent`: complete generated situation used to compute labels
|
| 48 |
+
- `provenance`: generator, split, renderer, and seed
|
| 49 |
+
|
| 50 |
+
## Generation
|
| 51 |
+
|
| 52 |
+
Generation code and manifests are included in `generator/` and `provenance/`. The policy makes labels mechanically identifiable: every component trial starts independently from the same original failing snapshot, and exactly one clearing intervention identifies the operational failure source.
|
| 53 |
+
|
| 54 |
+
## Limitations
|
| 55 |
+
|
| 56 |
+
- All scenarios are synthetic and templated.
|
| 57 |
+
- The label taxonomy covers only four component classes plus `unknown`.
|
| 58 |
+
- Recurring logical patterns across splits do not represent natural incident prevalence.
|
| 59 |
+
- Public test labels make the split unsuitable as a permanently sealed benchmark for future model selection.
|