Ultra-Innerthought

# Ultra-Innerthought🤔

[](https://huggingface.co/datasets/fnlp/Ultra-Innerthought)

<div align="left">
<a href="README.md">English</a> | <a href="README_zh.md">中文</a>
</div>

## Introduction
Ultra-Innerthought is a bilingual (Chinese and English) open-domain SFT dataset in Innerthought format, containing 2,085,326 dialogues. Unlike current reasoning datasets that mainly focus on mathematics and coding domains, Ultra-Innerthought covers a broader range of fields and includes both Chinese and English languages. We used Deepseek V3 as the model for data synthesis.

## Dataset Format
```json
{
"id": "dialogue_id",
"conversations": [
{
"user": "user_input",
"inner_thought": "model's inner thought",
"assistant": "model_output"
},
...
],
"data_source": "data_source"
}
```

## Data Synthesis
Ultra-Innerthought uses the following SFT datasets as raw input and employs Deepseek V3 for data synthesis. We preserved the user input from each round of the original datasets, using Deepseek V3 to first generate a model's Inner thought, and then generate the final response based on that Inner thought. When generating the model's Inner thought, we prompted the model to perform intent clarification, problem decomposition, self-reflection, exploration, and other behaviors. The dataset has approximately a 1:1 ratio of Chinese to English content.

### User Input Sources
User inputs are sampled from [OpenHerms2.5](https://huggingface.co/datasets/teknium/OpenHermes-2.5) and its Chinese version translated by Deepseek V3, [QwQ-LONGCOT-500K](https://huggingface.co/datasets/PowerInfer/QWQ-LONGCOT-500K) and its Chinese version translated by Deepseek V3, [tulu-3-sft-mixture](https://huggingface.co/datasets/allenai/tulu-3-sft-mixture), [sharegpt-zh](https://huggingface.co/datasets/kimnt93/zh-sharegpt), [COIG-CQIA](https://huggingface.co/datasets/m-a-p/COIG-CQIA), [Wildchat](https://huggingface.co/datasets/allenai/WildChat-1M), [WizardLM](https://huggingface.co/datasets/WizardLMTeam/WizardLM_evol_instruct_70k), Moss-inhouse-data, [lmsys](https://huggingface.co/datasets/lmsys/lmsys-chat-1m).

# Ultra-Innerthought🤔
[](https://huggingface.co/datasets/fnlp/Ultra-Innerthought)
<div align="left">
<a href="README.md">English</a> | <a href="README_zh.md">中文</a>
</div>

## 简介
Ultra-Innerthought是一个采用内部思考（Innerthought）格式的中英双语开放域监督微调（Supervised Fine-Tuning，SFT）数据集，共包含2,085,326条对话。相较于当前主要聚焦于数学与编码领域的推理数据集，Ultra-Innerthought覆盖了更广泛的领域，并同时支持中文与英文两种语言。我们采用Deepseek V3作为数据合成的基础模型。

## 数据集格式
json
{
"id": "dialogue_id",
"conversations": [
{
"user": "user_input",
"inner_thought": "model's inner thought",
"assistant": "model_output"
},
...
],
"data_source": "data_source"
}

## 数据合成流程
Ultra-Innerthought以如下SFT数据集作为原始输入，并借助Deepseek V3完成数据合成。我们保留了原始数据集每一轮的用户输入，先通过Deepseek V3生成模型的内部思考过程，再基于该思考过程生成最终回复。在生成模型内部思考时，我们提示模型执行意图澄清、问题拆解、自我反思、探索验证等行为。本数据集的中文与英文内容占比约为1:1。

### 用户输入来源
用户输入样本来源于以下数据集：[OpenHerms2.5](https://huggingface.co/datasets/teknium/OpenHermes-2.5) 及其由Deepseek V3翻译的中文版本、[QwQ-LONGCOT-500K](https://huggingface.co/datasets/PowerInfer/QWQ-LONGCOT-500K) 及其由Deepseek V3翻译的中文版本、[tulu-3-sft-mixture](https://huggingface.co/datasets/allenai/tulu-3-sft-mixture)、[sharegpt-zh](https://huggingface.co/datasets/kimnt93/zh-sharegpt)、[COIG-CQIA](https://huggingface.co/datasets/m-a-p/COIG-CQIA)、[Wildchat](https://huggingface.co/datasets/allenai/WildChat-1M)、[WizardLM](https://huggingface.co/datasets/WizardLMTeam/WizardLM_evol_instruct_70k)、Moss内部数据集、[lmsys](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)。