kristaller486/Nebo-T1-Russian

Nebo-T1-Russian数据集可能是第一个为俄语设计的longCoT类型的数据集，通过Deeseek-R1创建。提示信息来自于Sky-T1数据集，并通过Llama3.3-70B模型进行了翻译。回答和推理由Deeseek-R1生成（685B）。总共包含16.4K个样本，其中约12.4K个样本只包含俄语。回答和推论的语种使用fasttext进行了标注。部分样本缺少推论的标记，这些样本在`correct_format`列中被标记。

Nebo-T1-Russian is (probably) the first longCoT dataset for the Russian language created via Deeseek-R1. The prompts are taken from the Sky-T1 dataset and translated via Llama3.3-70B. The answers and reasoning are generated by Deeseek-R1 (685B). There are 16.4K samples in total, with approximately 12.4K being Russian-only. The languages in the answers and reasoning are labeled using fasttext. Some samples (≈400) are missing reasoning markup tags, and these samples are labeled in the `correct_format` column.