MMODELYST
Datasets/UltraFeedback
DAT

UltraFeedback

preference

Large-scale GPT-4-labeled preference dataset for RLHF/DPO.

Details
Kindpreference
Sourceopenbmb
Lifts these benchmarks
IFEval+6.0
BIG-Bench Hard+2.5
MMLU+1.5