y1001/Multilingual-Thinking

---
viewer: true
dataset_info:
features:
- name: reasoning_language
dtype: string
- name: developer
dtype: string
- name: user
dtype: string
- name: analysis
dtype: string
- name: final
dtype: string
- name: messages
list:
- name: content
dtype: string
- name: role
dtype: string
- name: thinking
dtype: string
splits:
- name: train
num_bytes: 8900623
num_examples: 1000
download_size: 5290171
dataset_size: 8900623
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
license: apache-2.0
task_categories:
- text-generation
language:
- en
- de
- fr
- es
- it
pretty_name: Multilingual-Thinking
size_categories:
- 1K<n<10K
---

# Dataset summary

Multilingual-Thinking is a reasoning dataset where the chain-of-thought has been translated from English into one of 4 languages: Spanish, French, Italian, and German. The dataset was created by sampling 1k training samples from the [SystemChat subset](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2/viewer/SFT/smoltalk_systemchats_Qwen3_32B_think) of [SmolTalk2](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2) and translating the reasoning traces with another language model.

This dataset was used in the [OpenAI Cookbook](https://cookbook.openai.com/articles/gpt-oss/fine-tune-transfomers) to fine-tune the OpenAI gpt-oss models.

You can load the dataset using:

```python
from datasets import load_dataset

ds = load_dataset("HuggingFaceH4/Multilingual-Thinking", split="train")
```

The `gpt-oss` models were trained on the Harmony response format for defining conversation structures, generating reasoning output and structuring function calls. The format is designed to mimic the OpenAI Responses API, and the table below summarizes the different message types used in the dataset:

| `developer` | The developer message is used to provide custom instructions for the model (what we usually call the `system` role) |
| :---- | :--|
| `user` | The user message is used to provide the input to the model |
| `assistant` | Output by the model which can either be a tool call or a message output. The output might also be associated with a particular “channel” identifying what the intent of the message is. |
| `analysis` | These are messages that are being used by the model for its chain-of thought |
| `final` | Messages tagged in the final channel are messages intended to be shown to the end-user and represent the responses from the model. |
| `messages` | The list of messages that combine the content of the above to produce a full conversation. This is the input to the model. |

If you're familiar with [OpenAI's messages format](https://platform.openai.com/docs/api-reference/messages/object), you will recognise this as being quite similar, but with an important difference:

> The `assistant` turn contains two special fields: a `thinking` one which contains the model's reasoning process, and a `content` one which contains the final response to the user.

查看器：已启用
数据集信息：
特征：
- 名称：reasoning_language，数据类型：字符串（string）
- 名称：developer，数据类型：字符串（string）
- 名称：user，数据类型：字符串（string）
- 名称：analysis，数据类型：字符串（string）
- 名称：final，数据类型：字符串（string）
- 名称：messages，列表类型：
- 子名称：content，数据类型：字符串（string）
- 子名称：role，数据类型：字符串（string）
- 子名称：thinking，数据类型：字符串（string）
划分集：
- 名称：train，字节数：8900623，样本数量：1000
下载大小：5290171，数据集总大小：8900623
配置项：
- 配置名称：default，数据文件：
- 划分：train，路径：data/train-*
许可协议：Apache-2.0
任务类别：文本生成（text-generation）
支持语言：英语（en）、德语（de）、法语（fr）、西班牙语（es）、意大利语（it）
易读名称：多语言思维（Multilingual-Thinking）
样本规模：1K < 样本数 <10K

# 数据集概述
多语言思维（Multilingual-Thinking）是一款推理类数据集，其思维链（Chain-of-Thought）已从英文翻译为西班牙语、法语、意大利语、德语这4种语言之一。本数据集的构建方式为：从[SmolTalk2](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2)的[SystemChat子集](https://huggingface.co/datasets/HuggingFaceTB/smoltalk2/viewer/SFT/smoltalk_systemchats_Qwen3_32B_think)中采样1000条训练样本，并通过另一款大语言模型（Large Language Model, LLM）对推理轨迹进行翻译。

本数据集曾被用于[OpenAI Cookbook](https://cookbook.openai.com/articles/gpt-oss/fine-tune-transfomers)，用于微调OpenAI的gpt-oss系列模型。

你可以通过以下代码加载该数据集：
python
from datasets import load_dataset

ds = load_dataset("HuggingFaceH4/Multilingual-Thinking", split="train")

gpt-oss系列模型基于Harmony响应格式进行训练，该格式用于定义对话结构、生成推理输出以及组织函数调用。此格式旨在模拟OpenAI Responses API，下表汇总了本数据集所使用的各类消息类型：

| 字段名 | 说明 |
| :---- | :--|
| `developer` | 开发者消息，用于为模型提供自定义指令（即我们通常所称的`system`角色） |
| `user` | 用户消息，用于向模型提供输入内容 |
| `assistant` | 模型生成的输出，可表现为工具调用或文本回复。该输出还可关联至特定“通道”，用于标识消息的意图 |
| `analysis` | 模型用于执行思维链（Chain-of-Thought）推理的消息 |
| `final` | 标记为`final`通道的消息，用于向终端用户展示，代表模型的最终回复 |
| `messages` | 整合上述各类消息以构成完整对话的消息列表，是模型的输入数据 |

如果你熟悉[OpenAI消息格式](https://platform.openai.com/docs/api-reference/messages/object)，会发现二者高度相似，但存在一处关键差异：
> `assistant`轮次包含两个特殊字段：`thinking`字段用于存储模型的推理过程，`content`字段用于存储向用户输出的最终回复。