Tralalabs/qa-dataset-436x-deduped

这是一个问答数据集，专为微调大型语言模型（LLMs）而设计。数据集通过LMArena的Battle Mode中的模型生成，使用提示语Write me a 65-row dataset with: question answer in JSONL for finetuning LLMs来创建。原始数据包含436行，经过去重处理后，剩余338行，移除了98行重复项。数据以JSONL格式组织，每行包含question（问题）和answer（答案）字段。

This is a question-and-answer dataset designed for fine-tuning large language models (LLMs). It was generated by models in Battle Mode on LMArena using the prompt Write me a 65-row dataset with: question answer in JSONL for finetuning LLMs. The dataset originally contained 436 rows, but after deduplication, it has 338 rows, with 98 rows removed as duplicates. The data is organized in JSONL format, with each row containing question and answer fields.