standrey/listing-descriptions-unsloth

listing_descriptions_dataset_full数据集由Unsloth Recipe Studio生成，包含1,654条生成记录。该数据集有2列，其中output列是LLM生成的文本内容，具有高唯一性(96.1%)。数据集采用NeMo Data Designer框架生成，该框架支持使用统计采样器、LLMs或现有种子数据集生成高质量合成数据，并能控制字段间的关系，进行质量验证和质量评估。The listing_descriptions_dataset_full dataset was developed by Unsloth Recipe Studio, comprising 1,654 generated records. This dataset includes two columns, where the `output` column stores text content generated by LLMs, with a uniqueness rate as high as 96.1%. The dataset was constructed using the NeMo Data Designer framework, which enables generating high-quality synthetic data via statistical samplers, LLMs, or existing seed datasets, supports controlling inter-field relationships, and facilitates both quality validation and assessment.