---
title: dureader-context-relevance-with-think
canonical_url: "https://www.modelscope.cn/datasets/zilliz/dureader-context-relevance-with-think"
md_url: "https://www.modelscope.cn/datasets/zilliz/dureader-context-relevance-with-think.md"
repository: zilliz/dureader-context-relevance-with-think
last_updated: 2026-01-16
license: other
storage_size: "216 MB"
downloads: 65
stars: 0
---

# dureader-context-relevance-with-think

> dureader-context-relevance-with-think - zilliz 在 ModelScope 开源的数据集。DuReader Context Relevance Dataset with Think Process

zilliz/dureader-context-relevance-with-think 是 ModelScope 魔搭社区上的数据集，存储大小 216 MB，采用 other 许可。

- **Repository**: zilliz/dureader-context-relevance-with-think
- **License**: other
- **Storage size**: 216 MB
- **Downloads**: 65
- **Stars**: 0
- **Last updated**: 2026-01-16

Source: https://www.modelscope.cn/datasets/zilliz/dureader-context-relevance-with-think

---

# DuReader Context Relevance Dataset with Think Process

This dataset is used for training the `zilliz/semantic-highlight-bilingual-v1`(https://huggingface.co/zilliz/semantic-highlight-bilingual-v1) model for semantic highlighting in RAG (Retrieval-Augmented Generation) systems.

## Dataset Description

This dataset contains query-context pairs with relevance annotations for context spans. The annotations help identify which parts of a document are semantically relevant to a query, even when they don't contain exact keyword matches.

### Key Features

- **`context_spans`**: Indicates the positions of segmented spans within the context text
- **`context_spans_relevance`**: Binary labels indicating whether each span should be highlighted (i.e., provides key information for answering the query)
- **`think_process`**: Contains the reasoning process used during annotation to ensure more accurate context span relevance labels and improve observability and interpretability

The dataset was annotated using Qwen3-8B, with the complete thinking process preserved in the `think_process` field.

## Data Source

This dataset is derived from `sentence-transformers/dureader`.

## License

This dataset follows the license of the original dataset.