Haakkim (حَكِّم, "judge") is an open arena-style human preference evaluation platform for Arabic large language models. This dataset contains all battle records collected during the first release snapshot, including voted comparisons and skipped battles across 11 Arabic dialect varieties.
Total… See the full description on the dataset page:
https://huggingface.co/datasets/Haakkim/Haakkim-1.0v.