Dataset · Ran0
JieZi ancient Chinese character VQA benchmark
Ran0/JieZi
A large expert VQA dataset for analyzing the evolution, etymology, and structure of ancient Chinese characters.
What's inside
Contains between 100,000 and 1 million examples in Chinese, linking high-quality glyph images with expert annotations. The labeling covers nine paleographical tasks, including uppercase character recognition, etymology, structural analysis, stroke evolution, and component functions.
What you can do with it
- Evaluating multimodal models (VLM) on complex ancient writing recognition and analysis tasks
- LoRA fine-tuning for OCR and paleographic analysis of Chinese texts
- Tasks
- Visual question answering, Question answering
- Languages
- Chinese
- Size
- 100K–1M rows
- Downloads this month
- 37.9K
- Likes
- 3
How to load it
from datasets import load_dataset
ds = load_dataset("Ran0/JieZi")