---
title: Human-Like-DPO-Dataset
canonical_url: "https://www.modelscope.cn/datasets/HumanLLMs/Human-Like-DPO-Dataset"
md_url: "https://www.modelscope.cn/datasets/HumanLLMs/Human-Like-DPO-Dataset.md"
repository: HumanLLMs/Human-Like-DPO-Dataset
last_updated: 2026-01-16
license: llama3
storage_size: "62 MB"
downloads: 3933
stars: 3
---

# Human-Like-DPO-Dataset

> Human-Like-DPO-Dataset - HumanLLMs 在 ModelScope 开源的数据集。Enhancing Human-Like Responses in Large Language Models

HumanLLMs/Human-Like-DPO-Dataset 是 ModelScope 魔搭社区上的数据集，存储大小 62 MB，采用 llama3 许可。

- **Repository**: HumanLLMs/Human-Like-DPO-Dataset
- **License**: llama3
- **Storage size**: 62 MB
- **Downloads**: 3933
- **Stars**: 3
- **Last updated**: 2026-01-16

Source: https://www.modelscope.cn/datasets/HumanLLMs/Human-Like-DPO-Dataset

---

<div align="center">
  <img src="https://cdn-avatars.huggingface.co/v1/production/uploads/63da3d7ae697e5898cb86854/H-vpXOX6KZu01HnV87Jk5.jpeg" width="320" height="320" />
  <h1>Enhancing Human-Like Responses in Large Language Models</h1>
</div>

<p align="center">
  🤗 <a href="https://huggingface.co/collections/HumanLLMs">Models</a> | 📊 <a href="https://huggingface.co/datasets/HumanLLMs/Human-Like-DPO-Dataset">Dataset</a> | 📄 <a href="https://arxiv.org/abs/2501.05032">Paper</a>
</p>

<p align="center">
  <b>📢 The paper associated with this dataset has been accepted to the <a href="https://personalizedllm.github.io/events/perfm-aaai26/">AAAI-26 Workshop on Personalization in the Era of Large Foundation Models (PerFM)</a>.</b>
</p>

# **Human-Like-DPO-Dataset**

This dataset was created as part of research aimed at improving conversational fluency and engagement in large language models. It is suitable for formats like **Direct Preference Optimization (DPO)** to guide models toward generating more human-like responses.

The dataset includes **10,884 samples** across **256 topics**, including:

- Technology
- Daily Life
- Science
- History
- Arts

Each sample contains:
- **Conversational Question**: Natural, engaging questions that reflect everyday human dialogue.
- **Human-Like Response**: A natural, conversational answer generated to mimic human interaction.
- **Formal Response**: A structured, professional answer reflecting traditional AI responses.

# **Dataset Usage**
This dataset can be used to fine-tune LLMs to:
- Improve conversational coherence.
- Reduce mechanical or impersonal responses.
- Enhance emotional intelligence in dialogue systems.

More details on dataset creation and usage can be found in the accompanying [research paper](https://arxiv.org/abs/2501.05032).
