README.md

---
language:
- fa
license: apache-2.0
size_categories:
- n<1K
task_categories:
- text-generation
- question-answering
tags:
- persian
- farsi
- reasoning
- thinking
- sft
- synthetic
pretty_name: Persian Thinking
---

# Persian-Thinking

**Persian-Thinking** is a small Persian-language reasoning/thinking dataset created by sampling and translating a subset of [SmolTalk2](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2).

## Dataset Details

- **1,000 samples** (999 after processing) drawn from the [`smoltalk_systemchats_Qwen3_32B_think`](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2/viewer/SFT/smoltalk_systemchats_Qwen3_32B_think) subset of SmolTalk2, part of its `SFT` split.
- That source subset consists of system-chat conversations generated with **Qwen3-32B in thinking mode**, meaning each assistant response includes an explicit reasoning trace before the final answer.
- All text (system prompt, user message, reasoning trace, and final response) was **translated into Persian using a language model**.
- The dataset preserves the original three-turn structure (system → user → assistant) but with all content localized to Persian.

## Dataset Creation

1. 1,000 rows were randomly sampled from the SmolTalk2 `smoltalk_systemchats_Qwen3_32B_think` subset (itself generated by HuggingFaceTB using Qwen3-32B on prompts from the original [SmolTalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk) systemchats-30k data).
2. Each sample's system prompt, user turn, reasoning trace (`thinking`), and final assistant response were translated to Persian with a language model.
3. The translated conversations were reshaped into flat columns (in addition to keeping the original nested `messages` structure) for easier downstream use in SFT pipelines.

No additional filtering, deduplication, or quality review beyond the source dataset's own decontamination was performed — see **Limitations** below.

## Dataset Structure

Each row contains:

| Column | Type | Description |
|---|---|---|
| `reasoning_language` | string | Always `"persian"` — the language of the `analysis` field. |
| `developer` | string | The system prompt (persona/instructions given to the assistant). |
| `user` | string | The user's message/question. |
| `analysis` | string | The assistant's Persian reasoning trace (chain-of-thought) before answering. |
| `final` | string | The assistant's final Persian response shown to the user. |
| `messages` | list | The original chat-formatted conversation (`system`, `user`, `assistant` turns), with the assistant turn carrying its `thinking` field separately from `content`. |

### Example

```json
{
  "reasoning_language": "persian",
  "developer": "تو یک دستیار هوش مصنوعی هستی که ...",
  "user": "می‌تونی کمکم کنی ...",
  "analysis": "خب، کاربر می‌خواد ...",
  "final": "حتماً ...",
  "messages": [
    {"role": "system", "content": "...", "thinking": null},
    {"role": "user", "content": "...", "thinking": null},
    {"role": "assistant", "content": "...", "thinking": "..."}
  ]
}
```

## Usage

```python
from datasets import load_dataset

ds = load_dataset("artindnr/Persian-Thinking", split="train")
print(ds[0]["analysis"])  # Persian reasoning trace
print(ds[0]["final"])     # Persian final answer
```

The `messages` column can be fed directly into chat-template-based SFT pipelines (e.g. TRL's `SFTTrainer`), while the flat `developer` / `user` / `analysis` / `final` columns are convenient for custom formatting, reasoning-specific training objectives, or inspection/filtering.

## Source Data

- **Base dataset:** [HuggingFaceTB/smoltalk2](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2), subset [`smoltalk_systemchats_Qwen3_32B_think`](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2/viewer/SFT/smoltalk_systemchats_Qwen3_32B_think) (`SFT` split), which HuggingFaceTB generated by running the prompts from [SmolTalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk) (systemchats-30k) through Qwen3-32B in reasoning mode.
- **Translation:** performed with a separate LLM, translating all conversation fields into Persian.

## Limitations

- The translation was performed by a language model and has not been manually reviewed; translation errors, awkward phrasing, or loss of nuance may be present.
- At 999 rows, this is a small sample intended for experimentation (e.g. quick fine-tuning trials, format prototyping) rather than large-scale training.
- Reasoning traces (`analysis`) reflect the original Qwen3-32B outputs translated to Persian, not reasoning generated natively in Persian — cultural/linguistic reasoning patterns may reflect English-first generation.
- No decontamination was independently re-verified for this subset beyond what SmolTalk2 already performed upstream.

## License

Inherits Apache 2.0 from the source `smoltalk-systemchats-Qwen3-32B` subset of SmolTalk2. Please also refer to the [SmolTalk2 license section](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2#license) for details on upstream components.

## Citation

If you use this dataset, please also credit the original SmolTalk2 dataset:

```bibtex
@misc{smoltalk2,
  author = {HuggingFaceTB},
  title = {SmolTalk2},
  howpublished = {\url{https://huggingface.co/datasets/HuggingFaceTB/smoltalk2}}
}
```