erjui/csrrg_ift_dataset

CSRRG指令微调数据集是一个大规模的指令遵循数据集，用于胸部X光报告生成。该数据集包含约160万个指令遵循示例，涵盖了结构化放射学报告生成（SRRG）和情境化结构化放射学报告生成（CSRRG）任务，包括发现和印象部分。数据集包含五个子集，分别为srrg_ift_dataset_findings、srrg_ift_dataset_impression、csrrg_ift_dataset_findings、csrrg_ift_dataset_impression和csrrg_ift_dataset_impression_indication_only。每个子集都包含训练、验证、测试和专家审核的测试数据。

CSRRG Instruction Fine-Tuning Dataset is a large-scale instruction-following dataset for chest X-ray report generation. The dataset contains approximately 1.6 million instruction-following examples across 5 subsets, covering both Structured Radiology Report Generation (SRRG) and Contextualized Structured Radiology Report Generation (CSRRG) tasks for FINDINGS and IMPRESSION sections. The dataset includes five subsets: srrg_ift_dataset_findings, srrg_ift_dataset_impression, csrrg_ift_dataset_findings, csrrg_ift_dataset_impression, and csrrg_ift_dataset_impression_indication_only, each with training, validation, test, and expert-reviewed test data.