Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- ps
|
| 5 |
+
- ar
|
| 6 |
+
- fa
|
| 7 |
+
- sd
|
| 8 |
+
- ur
|
| 9 |
+
tags:
|
| 10 |
+
- sft
|
| 11 |
+
- chat
|
| 12 |
+
- complex-reasoning
|
| 13 |
+
- multilingual
|
| 14 |
+
- pashto
|
| 15 |
+
- arabic
|
| 16 |
+
- farsi
|
| 17 |
+
- sindhi
|
| 18 |
+
- urdu
|
| 19 |
+
- synthetic-data
|
| 20 |
+
dataset_info:
|
| 21 |
+
features:
|
| 22 |
+
- name: messages
|
| 23 |
+
list:
|
| 24 |
+
- name: content
|
| 25 |
+
dtype: string
|
| 26 |
+
- name: role
|
| 27 |
+
dtype: string
|
| 28 |
+
- name: metadata
|
| 29 |
+
struct:
|
| 30 |
+
- name: difficulty
|
| 31 |
+
dtype: string
|
| 32 |
+
- name: domain
|
| 33 |
+
dtype: string
|
| 34 |
+
- name: license
|
| 35 |
+
dtype: string
|
| 36 |
+
- name: source_id
|
| 37 |
+
dtype: string
|
| 38 |
+
- name: subdomain
|
| 39 |
+
dtype: string
|
| 40 |
+
- name: task_type
|
| 41 |
+
dtype: string
|
| 42 |
+
- name: verification_status
|
| 43 |
+
dtype: string
|
| 44 |
+
---
|
| 45 |
+
|
| 46 |
+
# Pashto Plus Complex Reasoning Dataset
|
| 47 |
+
|
| 48 |
+
This dataset contains high-quality conversational and complex reasoning data tailored for Supervised Fine-Tuning (SFT) of Large Language Models (LLMs). It includes multi-turn chat interactions complete with system prompts, user instructions, assistant responses, and detailed metadata.
|
| 49 |
+
|
| 50 |
+
## Dataset Structure
|
| 51 |
+
|
| 52 |
+
Each record is formatted as a JSONL object containing chat messages and metadata:
|
| 53 |
+
|
| 54 |
+
```json
|
| 55 |
+
{
|
| 56 |
+
"messages": [
|
| 57 |
+
{
|
| 58 |
+
"content": "You are a rigorous research assistant...",
|
| 59 |
+
"role": "system"
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"content": "Your prompt or question here...",
|
| 63 |
+
"role": "user"
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"content": "Detailed reasoning and response here...",
|
| 67 |
+
"role": "assistant"
|
| 68 |
+
}
|
| 69 |
+
],
|
| 70 |
+
"metadata": {
|
| 71 |
+
"difficulty": "medium",
|
| 72 |
+
"domain": "biology",
|
| 73 |
+
"license": "apache-2.0",
|
| 74 |
+
"source_id": "domain-finetuning",
|
| 75 |
+
"subdomain": "laboratory_methods",
|
| 76 |
+
"task_type": "factual_explanation",
|
| 77 |
+
"verification_status": "unverified"
|
| 78 |
+
}
|
| 79 |
+
}
|
| 80 |
+
|
| 81 |
+
```
|
| 82 |
+
|
| 83 |
+
## Languages Included
|
| 84 |
+
|
| 85 |
+
* **Pashto (ps)**
|
| 86 |
+
* **Arabic (ar)**
|
| 87 |
+
* **Persian / Farsi (fa)**
|
| 88 |
+
* **Sindhi (sd)**
|
| 89 |
+
* **Urdu (ur)**
|
| 90 |
+
|
| 91 |
+
## License
|
| 92 |
+
|
| 93 |
+
Distributed under the **Apache-2.0** license.
|
| 94 |
+
|
| 95 |
+
```
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
```
|