CCFQA

CCFQA is a speech and text factuality evaluation benchmark that measures language models’ ability to answer short, fact-seeking questions and assess their cross-lingual and cross-modal consistency. It consists of speech and text in 8 languages, containing 1,800 n-way parallel sentences and a total of 14,000 speech samples.

• Language: Mandarin Chinese, English, French, Japanese, Korean, Russian, Spanish, Cantonese(HK)
• ISO-3 Code: cmn, eng, fra, jpn, kor, rus, spa, yue

📄Paper：https://arxiv.org/abs/2508.07295

How to use

from datasets import load_dataset

ccfqa = load_dataset("yxdu/ccfqa")
print(ccfqa)

⚖️ Evals

please visit github page.

License

All datasets are licensed under the Creative Commons Attribution-NonCommercial license (CC-BY-NC), which allows use, sharing, and adaptation for non-commercial purposes only, with proper attribution.

🖊Citation

@misc{du2025ccfqabenchmarkcrosslingualcrossmodal,
      title={CCFQA: A Benchmark for Cross-Lingual and Cross-Modal Speech and Text Factuality Evaluation}, 
      author={Yexing Du and Kaiyuan Liu and Youcheng Pan and Zheng Chu and Bo Yang and Xiaocheng Feng and Yang Xiang and Ming Liu},
      year={2025},
      eprint={2508.07295},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2508.07295}, 
}