bombman/thwiki-2026-super-clean-2k

该数据集是从泰语维基百科（2026年Dump）中提取的，专注于Pro-Clean级别的清洁度，专门用于大语言模型（LLM）的蒸馏和微调。它具有高信息密度，仅选择长度超过1000个字符且泰语内容比例超过50%的文章；经过专业清洁，移除了图像文件名（如.jpg、.png）、维基符号垃圾（如==, ""）和空括号；同时保留引号（" "）以维持专有名词和地点的边界；并通过从整个文件中随机抽取数据来确保内容的多样性。This dataset is extracted from the Thai Wikipedia (Dump 2026) with a focus on Pro-Clean level cleanliness, specifically designed for distillation and fine-tuning of LLMs. It features high information density by selecting only articles longer than 1,000 characters with a Thai language proportion > 50%; is pro-cleaned by removing image file names (.jpg, .png), wiki symbol clutter (==, "), and empty brackets; preserves quotation marks (" ") to maintain the boundaries of proper nouns and places; and is randomized by sampling data from the entire file to ensure content diversity.