Datasets:
language:
- en
- zh
license: cc-by-nc-2.0
size_categories:
- 1K<n<10K
task_categories:
- visual-question-answering
- image-to-image
tags:
- culture
- hanfu
Dataset Card for Hanfu-Bench
Official Dataset for Hanfu-Bench: A Multimodal Benchmark on Cross-Temporal Cultural Understanding and Transcreation
Dataset Description
We introduce Hanfu-Bench, a novel, expert-curated multimodal dataset. Hanfu, a traditional garment spanning ancient Chinese dynasties, serves as a representative cultural heritage that reflects the profound temporal aspects of Chinese culture while remaining highly popular in Chinese contemporary society. Hanfu-Bench comprises two core tasks: cultural visual understanding and cultural image transcreation. The former task examines temporal-cultural feature recognition based on single- or multi-image inputs through multiple-choice visual question answering, while the latter focuses on transforming traditional attire into modern designs through cultural element inheritance and modern context adaptation. Our evaluation shows that closed VLMs perform comparably to non-experts on visual cutural understanding but fall short by 10% to human experts, while open VLMs lags further behind non-experts. For the transcreation task, multi-faceted human evaluation indicates that the best-performing model achieves a success rate of only 42%. Our benchmark provides an essential testbed, revealing significant challenges in this new direction of temporal cultural understanding and creative adaptation.
Usage
This dataset contains two json files containing the single-image VQA task, and multi-image VQA task, accompanied by the corresponding images.
Citation
@misc{zhou2025hanfubenchmultimodalbenchmarkcrosstemporal,
title={Hanfu-Bench: A Multimodal Benchmark on Cross-Temporal Cultural Understanding and Transcreation},
author={Li Zhou and Lutong Yu and Dongchu Xie and Shaohuan Cheng and Wenyan Li and Haizhou Li},
year={2025},
eprint={2506.01565},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://huggingface.co/papers/2506.01565},
}
Dataset Card Authors
Li Zhou, Lutong Yu, Dongchu Xie, Shaohuan Cheng, Wenyan Li, Haizhou Li
Dataset Card Contact
If you have any questions regarding the dataset, please reach out to: lizhou21@cuhk.edu.cn