barandinho/turkish-reasoning-distilled-sft

该数据集包含土耳其语推理SFT数据，用于模型barandinho/qwen3.5-27b-tudum-dapo-50的监督微调。数据结合了来自多个来源的生成土耳其语推理轨迹和已验证的教师-SFT轨迹，包括DAPO数学、WebInstruct、AceCode、IFEval风格指令遵循、OpenThoughts科学、OpenThoughts代码和系统聊天数据。数据集包含多个配置，如all、verified等，每个配置对应不同的数据子集。数据集的schema包括system_prompt、user_prompt、reasoning_content、final_answer、source、gen_id和verified字段。总行数为411,657，验证行数为292,458，总体验证比例为71.04%。还提供了各来源的token统计，如平均提示token数、平均输出token数等。

This dataset contains Turkish reasoning SFT data for the model barandinho/qwen3.5-27b-tudum-dapo-50. It combines generated Turkish reasoning traces from DAPO math, WebInstruct, AceCode, and verifier-compatible IFEval-style instruction-following sources with verified teacher-SFT traces from DAPO math, OpenThoughts science, OpenThoughts code, and system-chat data. The dataset includes multiple configurations such as all, verified, etc., each corresponding to different data subsets. The schema includes fields like system_prompt, user_prompt, reasoning_content, final_answer, source, gen_id, and verified. Total rows are 411,657, verified rows are 292,458, and the overall verification ratio is 71.04%. Token statistics for each source are also provided, including mean prompt tokens, mean output tokens, etc.