Add dataset card
Browse files
README.md
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pretty_name: MatrAIx Wiki Personas
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
size_categories:
|
| 6 |
+
- 10K<n<100K
|
| 7 |
+
---
|
| 8 |
+
# MatrAIx Wiki Personas
|
| 9 |
+
|
| 10 |
+
This repository currently contains one complete sample shard of persona attributes extracted from English Wikipedia-derived public-person profiles using `Qwen/Qwen3.6-35B-A3B`.
|
| 11 |
+
|
| 12 |
+
## Files
|
| 13 |
+
|
| 14 |
+
- `data/shard_0001.jsonl.gz`: **10,630 personas** (one JSON object per line), gzip-compressed from a 2.7 GB JSONL file.
|
| 15 |
+
- `schema/dimensions_legacy_v1.json`: the exact legacy persona dimension schema used when this shard was extracted.
|
| 16 |
+
|
| 17 |
+
Compressed shard SHA-256: `d3765cba51934f8aa765fce9dc9df6d31af4b0ad34181e8345bfeaa82bd15c51`.
|
| 18 |
+
|
| 19 |
+
## Record format
|
| 20 |
+
|
| 21 |
+
```json
|
| 22 |
+
{
|
| 23 |
+
"global_idx": 10630,
|
| 24 |
+
"qid": "Q...",
|
| 25 |
+
"title": "Person name",
|
| 26 |
+
"fields": [{
|
| 27 |
+
"field_id": "age_bracket",
|
| 28 |
+
"value": "35-44",
|
| 29 |
+
"confidence": 1.0,
|
| 30 |
+
"evidence": "verbatim source quote",
|
| 31 |
+
"description": "1-2 sentence person-specific attribute description",
|
| 32 |
+
"assignment_type": "direct"
|
| 33 |
+
}]
|
| 34 |
+
}
|
| 35 |
+
```
|
| 36 |
+
|
| 37 |
+
Each persona has approximately 1,290 field objects. Values and descriptions are model-generated inferences grounded, where available, by evidence quotes. They may contain errors and should not be treated as verified biographical facts.
|
| 38 |
+
|
| 39 |
+
Source dataset: [`MatrAIx2026/MatrAIx2026`](https://huggingface.co/datasets/MatrAIx2026/MatrAIx2026).
|