introvoyz041/synthvision-annotated-qwen

该数据集名为synthvision-annotated-qwen，包含59,476条医疗图像标注记录，这些记录是SynthVision流程的前半部分标注，由Qwen 3.5（397B MoE，17B激活）模型通过Doubleword批量推理生成。每个记录包括多轮临床对话（5-9轮）、临床叙述报告（4-6句话）、结构化发现（以字典形式表示，键为发现名称，值为对应值）、推理链（编号的诊断步骤）和难度评级（简单、中等或困难）。数据集旨在支持视觉问答任务，特别是在医疗领域，用于模拟临床诊断和报告生成。

The dataset is named synthvision-annotated-qwen and consists of 59,476 medical image annotations, which are the first-half annotations from the SynthVision pipeline. These annotations were generated by the Qwen 3.5 (397B MoE, 17B active) model via Doubleword batch inference. Each record contains a multi-turn clinical conversation (5-9 turns), a clinical narrative report (4-6 sentences), structured findings (in dictionary format with finding_name → value), a reasoning chain (numbered diagnostic steps), and a difficulty rating (easy, intermediate, or hard). The dataset is designed for visual question answering tasks, particularly in the medical domain, to simulate clinical diagnosis and report generation.