WAON: Large-Scale and High-Quality Japanese Image-Text Pair Dataset for Vision-Language Models
WAON is a Japanese (image, text) pair dataset containing approximately 155M examples, crawled from Common Crawl.
It is built from snapshots taken in 2025-18, 2025-08, 2024-51, 2024-42, 2024-33, and 2024-26.
The dataset is high-quality and diverse, constructed through a sophisticated… See the full description on the dataset page:
https://huggingface.co/datasets/speed/WAON.
