Full Standardized Prompt for Vision–Language Model Data Extraction (hema_v6)

This dataset contains the complete standardized prompt used to instruct vision–language models (VLMs) such as Qwen 2.5, Gemini 2.5 Pro, and GPT-5 in the extraction of structured information from scanned hematology and hemogram laboratory reports. The prompt defines strict schema conformity rules, type guards, and validation constraints to ensure consistency between model outputs. It specifies detailed formatting, normalization, and data-handling instructions for all fields, including numeric values, units, reference ranges, and report metadata.本数据集包含用于指导视觉语言模型（vision–language models，VLMs）——如Qwen 2.5、Gemini 2.5 Pro及GPT-5——从扫描版血液学与血常规实验室报告中提取结构化信息的完整标准化提示词。该提示词定义了严格的模式一致性规则、类型守卫机制与验证约束，以确保模型输出的一致性。它针对所有字段（包括数值、单位、参考范围及报告元数据）规定了详尽的格式化、归一化与数据处理要求。