You're currently viewing an old version of this dataset. To see the current version, click here.

UltraFeedback

UltraFeedback is a dataset consisting of AI feedback where GPT-4 rates responses obtained from four different language models for the 64k prompts, based on four principles of helpfulness, instruction-following, truthfulness, and honesty.

Data and Resources

Cite this as

Ganqu Cui, Lifan Yuan, Ning Ding, Guanming Yao, Wei Zhu, Yuan Ni, Guotong Xie, Zhiyuan Liu, Maosong Sun (2024). Dataset: UltraFeedback. https://doi.org/10.57702/06vp56te

DOI retrieved: December 16, 2024

Additional Info

Field Value
Created December 16, 2024
Last update December 16, 2024
Defined In https://doi.org/10.48550/arXiv.2402.11253
Citation
  • https://doi.org/10.48550/arXiv.2407.06645
Author Ganqu Cui
More Authors
Lifan Yuan
Ning Ding
Guanming Yao
Wei Zhu
Yuan Ni
Guotong Xie
Zhiyuan Liu
Maosong Sun
Homepage https://huggingface.co/datasets/openbmb/UltraFeedback