introvoyz041/synthvision-validated-qwen-by-kimi

该数据集名为synthvision-validated-qwen-by-kimi，是一个医疗领域的视觉问答数据集，属于SynthVision流程的交叉验证子集。数据集包含55,359条记录，由Kimi K2.5对59,476个Qwen 3.5注释进行审查，确认了其中93.1%的注释与源图像一致。验证标准要求注释一致（consistent == true）且置信度不低于0.7（confidence >= 0.7），未通过验证的记录（主要是标注者幻觉出图像中未见发现的情况）已被移除。数据模式包括唯一ID、图像路径、多轮对话（ShareGPT格式）、临床报告、结构化发现、验证信息和综合质量评分。数据集规模在10K到100K之间，适用于医疗视觉问答任务。This dataset is named synthvision-validated-qwen-by-kimi, a cross-validated subset from the SynthVision pipeline. Kimi K2.5 reviewed all 59,476 Qwen 3.5 annotations and confirmed 55,359 as consistent with the source images, resulting in a 93.1% pass rate. Validation criteria required consistent == true AND confidence >= 0.7, and records that failed validation (primarily cases where the annotator hallucinated findings not visible in the image) were removed. The schema includes id, image path, conversations in multi-turn ShareGPT format, clinical report, structured findings, validation details, and a composite quality score. It is categorized under visual-question-answering with medical, synthvision, and openmed tags, and has a size category of 10K to 100K records.