Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DanQing: An Up-to-Date Large-Scale Chinese Vision-Language Pre-training Dataset

Hengyu Shen, Tiancheng Gu, Bin Qin, Lan Wu, Yuling Wu, Shuo Tan, Zelong Sun, Jun Wang, Nan Wu, Xiang An, Weidong Cai, Ziyong Feng, Kaicheng Yang

TL;DR

The proposed DanQing dataset, which contains 100 million image-text pairs collected from Common Crawl, is presented, which consistently achieves superior performance across a range of Chinese downstream tasks, including zero-shot classification, cross-modal retrieval, and LMM-based evaluations.

Abstract

Vision-Language Pre-training (VLP) models demonstrate strong performance across various downstream tasks by learning from large-scale image-text pairs through contrastive pretraining. The release of extensive English image-text datasets (e.g., COYO-700M and LAION-400M) has enabled widespread adoption of models such as CLIP and SigLIP in tasks including cross-modal retrieval and image captioning. However, the advancement of Chinese vision-language pretraining has substantially lagged behind, due to the scarcity of high-quality Chinese image-text data. To address this gap, we develop a comprehensive pipeline for constructing a high-quality Chinese cross-modal dataset. As a result, we propose DanQing, which contains 100 million image-text pairs collected from Common Crawl. Different from existing datasets, DanQing is curated through a more rigorous selection process, yielding superior data quality. Moreover, DanQing is primarily built from 2024-2025 web data, enabling models to better capture evolving semantic trends and thus offering greater practical utility. We compare DanQing with existing datasets by continual pre-training of the SigLIP2 model. Experimental results show that DanQing consistently achieves superior performance across a range of Chinese downstream tasks, including zero-shot classification, cross-modal retrieval, and LMM-based evaluations. To facilitate further research in Chinese vision-language pre-training, we will open-source the DanQing dataset under the Creative Common CC-BY 4.0 license.

DanQing: An Up-to-Date Large-Scale Chinese Vision-Language Pre-training Dataset

TL;DR

The proposed DanQing dataset, which contains 100 million image-text pairs collected from Common Crawl, is presented, which consistently achieves superior performance across a range of Chinese downstream tasks, including zero-shot classification, cross-modal retrieval, and LMM-based evaluations.

Abstract

Vision-Language Pre-training (VLP) models demonstrate strong performance across various downstream tasks by learning from large-scale image-text pairs through contrastive pretraining. The release of extensive English image-text datasets (e.g., COYO-700M and LAION-400M) has enabled widespread adoption of models such as CLIP and SigLIP in tasks including cross-modal retrieval and image captioning. However, the advancement of Chinese vision-language pretraining has substantially lagged behind, due to the scarcity of high-quality Chinese image-text data. To address this gap, we develop a comprehensive pipeline for constructing a high-quality Chinese cross-modal dataset. As a result, we propose DanQing, which contains 100 million image-text pairs collected from Common Crawl. Different from existing datasets, DanQing is curated through a more rigorous selection process, yielding superior data quality. Moreover, DanQing is primarily built from 2024-2025 web data, enabling models to better capture evolving semantic trends and thus offering greater practical utility. We compare DanQing with existing datasets by continual pre-training of the SigLIP2 model. Experimental results show that DanQing consistently achieves superior performance across a range of Chinese downstream tasks, including zero-shot classification, cross-modal retrieval, and LMM-based evaluations. To facilitate further research in Chinese vision-language pre-training, we will open-source the DanQing dataset under the Creative Common CC-BY 4.0 license.
Paper Structure (24 sections, 1 equation, 11 figures, 7 tables)

This paper contains 24 sections, 1 equation, 11 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Vision-Language Pretraining
  4. Large-Scale Image-Text Dataset
  5. DanQing Dataset
  6. Training Objective of DanQing
  7. Curation of DanQing
  8. Statistic of DanQing
  9. Experiments and Results
  10. Implementation Details
  11. Main Results
  12. Analysis
  13. Text Quality Analysis
  14. Scaling Capability
  15. Image Semantic Balance
  16. ...and 9 more sections

Figures (11)

  • Figure 1: Overview of the DanQing dataset construction pipeline.
  • Figure 2: Overview of data characteristics in DanQing.
  • Figure 3: Visualization of popular topic in the DanQing dataset, generated via BERTopic grootendorst2022bertopicneuraltopicmodeling on 10M subset.
  • Figure 4: Text Quality Analysis. Randomly sample 10M subsets from DanQing, Wukong, and Zero, then compare their content word density and perplexity.
  • Figure 5: Data scaling and model scaling capability comparison between DanQing and Wukong.
  • ...and 6 more figures