File size: 4,884 Bytes
6873d1f
1cee455
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cc0ba9c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6873d1f
1cee455
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
---
language:
- tr
license: mit
task_categories:
- text-generation
- question-answering
tags:
- turkish
- identity
- instruction-tuning
- llm-alignment
- nlp
- chatbot
pretty_name: TurkishIdentityMini
size_categories:
- n<1K
dataset_info:
  features:
  - name: instruction
    dtype: string
  - name: output
    dtype: string
  splits:
  - name: train
    num_bytes: 41540
    num_examples: 481
  download_size: 13573
  dataset_size: 41540
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# TurkishIdentityMini

## Dataset Description

**TurkishIdentityMini** is a small, template-based Turkish instruction dataset designed to help LLMs respond correctly to identity-related questions. It contains instruction–output pairs where a user asks a chatbot about its name, origin, or creator, and the model responds using customizable `{{model_name}}` and `{{team_name}}` placeholders.

This dataset is useful for fine-tuning or instruction-tuning Turkish language models to maintain a consistent, branded identity and to correctly deny affiliation with other AI providers such as OpenAI, Google, Meta, Anthropic, or Microsoft.

### Dataset Summary

| Property | Value |
|---|---|
| Language | Turkish (`tr`) |
| Split | `train` only |
| Format | Parquet |
| License | MIT |

---

## Dataset Structure

### Data Fields

| Field | Type | Description |
|---|---|---|
| `instruction` | `string` | A Turkish user query about the model's identity (e.g., *"Sen kimsin?"*, *"Seni kim yaptı?"*) |
| `output` | `string` | A template response using `{{model_name}}` and `{{team_name}}` placeholders |

### Example Rows

| instruction | output |
|---|---|
| `Seni kim yaptı?` | `Ben {{team_name}} ekibi tarafından yapıldım.` |
| `Sen kimsin?` | `Ben {{model_name}}, {{team_name}} tarafından geliştirilmiş bir yapay zeka asistanıyım.` |
| `ChatGPT misin?` | `Hayır, ben {{model_name}}. {{team_name}} tarafından eğitilmiş bir asistanım.` |
| `OpenAI'dan mısın?` | `Hayır, ben {{team_name}} tarafından bağımsız olarak geliştirildim.` |
| `Adın ne?` | `Ben {{model_name}}, yapay zeka asistanınızım.` |

---

## Intended Uses

### Primary Use

Fine-tuning or instruction-tuning Turkish LLMs to:
- Respond to identity queries (name, creator, affiliation) in Turkish
- Correctly deny association with third-party AI providers
- Use a consistent, customizable branded persona via the `{{model_name}}` and `{{team_name}}` placeholders

### Out-of-Scope Use

- This dataset covers **only** identity-related queries; it is not suitable as a standalone fine-tuning corpus for general conversational ability
- The placeholder format requires preprocessing before use in most training pipelines

---

## Dataset Creation

### Covered Question Categories

The dataset covers the following identity query themes:

- **Creator / origin***"Seni kim yaptı?"*, *"Nereden geliyorsun?"*
- **Name / model identity***"Adın ne?"*, *"Model adını söyler misin?"*
- **Brand denial***"ChatGPT misin?"*, *"Sen Claude musun?"*, *"Google tarafından mı oluşturuldun?"*
- **Greetings with identity***"Merhaba"*, *"Selam"* → model introduces itself
- **Paraphrastic variants** — Diverse rephrasings of the same intents to improve robustness

### Template Placeholders

All outputs use two placeholders that must be filled before training:

| Placeholder | Description |
|---|---|
| `{{model_name}}` | The name of the deployed model |
| `{{team_name}}` | The name of the developing team or organization |

**Example preprocessing (Python):**
```python
def fill_template(example, model_name, team_name):
    example["output"] = (
        example["output"]
        .replace("{{model_name}}", model_name)
        .replace("{{team_name}}", team_name)
    )
    return example

dataset = dataset.map(lambda x: fill_template(x, "Magibu-11b-v0.8", "magibu"))
```

---

## Usage

### With 🤗 Datasets

```python
from datasets import load_dataset

dataset = load_dataset("aliarda/TurkishIdentityMini")
print(dataset["train"][0])
# {'instruction': 'Seni kim yaptı?', 'output': 'Ben {{team_name}} ekibi tarafından yapıldım.'}
```

### With pandas

```python
import pandas as pd

df = pd.read_parquet("hf://datasets/aliarda/TurkishIdentityMini/data/train-*.parquet")
print(df.head())
```

---

## Acknowledgements

80 rows in this dataset were sourced from [`sts07142/llm-name-identity`](https://huggingface.co/datasets/sts07142/llm-name-identity) and translated into Turkish using AI-assisted translation.

---

## Citation

If you use this dataset in your research, please cite it as:

```bibtex
@dataset{aliarda_turkishidentitymini,
  author    = {Ali Arda Fincan},
  title     = {TurkishIdentityMini},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/aliarda/TurkishIdentityMini}
}
```