dashb0ardtech/master-cotizador-v3-dataset

Master Cotizador v3数据集用于微调模型，以自动生成加油站维护服务的报价。数据集包含9,736个对话示例，用于训练助手根据报告的问题/损坏、维护类型（纠正性、预防性、校准）以及加油站和石油公司生成维护报价。每个示例遵循对话格式：系统提供专业助手的指令，用户提供包含问题细节的报价请求，助手提供包含诊断、解决方案和材料的完整报价。数据集统计显示，总示例数为9,736，训练集8,762，验证集974，平均令牌数约为1,500，71.5%的示例包含材料。维护类型分布为：纠正性65.3%，校准15.2%，预防性12.6%，其他6.9%。石油公司分布为：DELTA 94%，TEXACO、ULTRA、MIT、PUMA、TERPEL 6%。

The Master Cotizador v3 Dataset is designed for fine-tuning models to automatically generate maintenance service quotations for gas stations. The dataset contains 9,736 conversation examples to train an assistant that generates maintenance quotations based on: the reported problem/damage, the type of maintenance (Corrective, Preventive, Calibration), and the station and oil company. Each example follows a conversational format: System provides instructions for the specialized assistant, User requests a quotation with problem details, and Assistant provides a complete quotation with diagnosis, solution, and materials. Statistics show the total examples are 9,736, with 8,762 for training and 974 for validation, averaging ~1,500 tokens, and 71.5% include materials. Maintenance type distribution is: CORRECTIVO 65.3%, CALIBRACION 15.2%, PREVENTIVO 12.6%, OTRO 6.9%. Oil company distribution is: DELTA 94%, TEXACO, ULTRA, MIT, PUMA, TERPEL 6%.