---
title: Drone-flight-monitoring-reasoning-SFT
canonical_url: "https://www.modelscope.cn/datasets/gabrielai/Drone-flight-monitoring-reasoning-SFT"
md_url: "https://www.modelscope.cn/datasets/gabrielai/Drone-flight-monitoring-reasoning-SFT.md"
repository: gabrielai/Drone-flight-monitoring-reasoning-SFT
chinese_name: "无人机飞行安全中文问答数据集"
last_updated: 2025-04-10
license: "Apache License 2.0"
storage_size: "4.6 MB"
downloads: 274
stars: 6
---

# Drone-flight-monitoring-reasoning-SFT

> Drone-flight-monitoring-reasoning-SFT - gabrielai 在 ModelScope 开源的数据集。本数据集是一个专注于**无人机飞行安全**领域的**中文问答数据集**，采用了**Chain-of-Thought (CoT)** 的格式。它旨在用于练习大语言模型的微调训练，使其能够模拟专家思考过程，并针对无人机安全相关问题生成包含推理步骤的结构化回答。

gabrielai/Drone-flight-monitoring-reasoning-SFT 是 ModelScope 魔搭社区上的数据集，存储大小 4.6 MB，采用 Apache License 2.0 许可。

- **Repository**: gabrielai/Drone-flight-monitoring-reasoning-SFT
- **License**: Apache License 2.0
- **Storage size**: 4.6 MB
- **Downloads**: 274
- **Stars**: 6
- **Last updated**: 2025-04-10

Source: https://www.modelscope.cn/datasets/gabrielai/Drone-flight-monitoring-reasoning-SFT

---

# Drone-flight-monitoring-reasoning-SFT

## Dataset Description

本数据集是一个专注于**无人机飞行安全**领域的**中文问答数据集**，采用了**Chain-of-Thought (CoT)** 的格式。它旨在用于练习大语言模型的微调训练，使其能够模拟专家思考过程，并针对无人机安全相关问题生成包含推理步骤的结构化回答。微调后模型见([gabrielai/Deepseek-r1-finetuned-drone-safty](https://www.modelscope.cn/models/gabrielai/Deepseek-r1-finetuned-drone-safty)) 。

本数据集是基于 Hugging Face 平台上的 `skylink-drone-cot-datasets` ([pohsjxx/default-domain-cot-dataset](https://huggingface.co/datasets/pohsjxx/default-domain-cot-dataset)) 进行处理和衍生的。



## Dataset Structure / Data Fields

数据集中的每个样本包含以下字段：

*   `Question` (string): 关于无人机飞行安全或风险相关的问题。
*   `Reasoning` (string): 模拟模型的推理过程。
*   `Answer` (string): 模型的最终正式回答。



## Dataset Creation

### Source Data

*   **Original Dataset Name:** skylink-drone-cot-datasets
*   **Original Dataset Link:** https://huggingface.co/datasets/pohsjxx/default-domain-cot-dataset
*   **Original Author/Provider:** pohsjxx
*   **Original License:** Apache License 2.0

### Processing Steps

本数据集在原始 `skylink-drone-cot-datasets` 的基础上进行了以下主要处理步骤：

1.  **Data Extraction:** 从原始数据集中提取了约 3500 个的样本。
2.  **Format Alignment:** 调整了原数据集中 answer 字段的文字表述，使其更通用。
3.  **Label Exchange:** 为了适配 Chain-of-Thought (CoT) 的回答模式，并使其更符合某些预训练模型（如 DeepSeek 系列）的微调数据格式，对原始数据集中的 `reasoning` 和 `answer` 字段进行了**对调处理**。具体来说：
    *   原始数据集中的 `answer` 字段内容被用作本数据集中回答的“思考过程”部分（ `</think>` 标签前的部分）。
    *   原始数据集中的 `reasoning` 字段内容被用作本数据集中回答的“最终答案”部分（ `</think>` 标签后的部分）。


## Usage

你可以使用如下代码加载本数据集：

```python
from modelscope.msdatasets import MsDataset
ds =  MsDataset.load('gabrielai/Drone-flight-monitoring-reasoning-SFT')
