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

TL;DR

DanQing is a comprehensive large-scale Chinese vision-language dataset with 100 million high-quality image-text pairs, curated through a systematic pipeline, enabling improved model performance on various downstream tasks and capturing contemporary semantic trends.

Contribution

The paper introduces DanQing, a new large-scale Chinese cross-modal dataset with advanced data curation techniques, addressing the lack of high-quality open-source data for Chinese VLP.

Findings

DanQing outperforms existing Chinese datasets in downstream tasks.

The dataset captures recent semantic trends from 2024-2025.

It exhibits a balanced semantic distribution and better scaling capability.

Abstract

Vision-Language Pre-training (VLP) models have achieved remarkable success by leveraging large-scale image-text pairs. While English-centric models like CLIP and SigLIP benefit from massive datasets (e.g., LAION-400M), the development of Chinese VLP remains bottlenecked by the lack of high-quality, large-scale open-source data. In this paper, we present DanQing, a large-scale Chinese cross-modal dataset containing 100 million high-quality image-text pairs curated from Common Crawl. To ensure superior data quality, we develop an effective systematic pipeline comprising data source selection, text refinement, visual diversification, and cross-modal cross-batch filtering, thereby effectively mitigating the intrinsic noise prevalent in web data. Notably, DanQing incorporates data from 2024-2025, enabling models to capture contemporary semantic trends and emerging concepts. Extensive experiments via continued pretraining of SigLIP2 models demonstrate that DanQing consistently outperforms existing Chinese datasets across diverse downstream tasks, including zero-shot classification, cross-modal retrieval, and Chinese-centric large multimodal model tasks. Furthermore, in-depth analysis of DanQing reveals that it exhibits a more balanced semantic distribution and superior scaling capability compared to existing datasets. To facilitate further research in Chinese vision-language pre-training, we will open-source the DanQing dataset under the Creative Common CC-BY-NC 4.0 license.