mithraindir/macrostories-micro-2layer-d48
Text Generation • 71.6k • Updated • 54
Datasets:
Dataset Viewer
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
MacroStories Stories Dataset
Synthetic corpus of strictly structured Dutch children's stories generated and validated for the MacroStories research project.
Overview
The dataset provides certified 4-beat Dutch micro-stories centered around a compatible object-obstacle pairing matrix:
- ****: 9,856 certified consistent Dutch stories (~200 words each).
- ****: 5,544 distilled 4-sentence micro-arcs (~75 words each) with explicit structural boundary tokens (, , , ).
- ****: Standalone 800-token Byte-Level BPE tokenizer matching the vocabulary of MacroStories models (0.00% OOV on Dutch morphology).
Key Characteristics
- Strict Narrative Consistency: All stories enforce a primary object introduced in sentence 1 and retained through the resolution.
- Compatible Scenario Matrix: Pairs logical tools with sensible hurdles (e.g., hammer $\to$ wobbly bench/birdhouse, bucket $\to$ drought/sandcastle) to prevent hallucinated competing tools.
- De-anonymized Entities: Uses placeholders easily mappable to regional names.
- Downloads last month
- 40
Models trained or fine-tuned on mithraindir/macrostories-stories