qikp/legal-sft

---
task_categories:
- text-generation
language:
- en
tags:
- legal
---

# legal-sft

This is a dataset, derived from [nguha/legalbench](https://huggingface.co/datasets/nguha/legalbench), and is formatted to be a SFT dataset. Only rows with `question`, and that don't include `text` (i.e. RAG) are kept. Conversations containing "Consider the country" are removed.

---
任务类别：
- 文本生成（text-generation）
语言：
- 英语（en）
标签：
- 法律（legal）
---

# legal-sft

本数据集源自[nguha/legalbench](https://huggingface.co/datasets/nguha/legalbench)，并被格式化为监督微调（Supervised Fine-Tuning，简称SFT）数据集。仅保留包含`question`字段且未包含`text`（即检索增强生成（Retrieval-Augmented Generation，简称RAG））的样本行。所有包含"Consider the country"的对话样本均已移除。