MIM-VQA
收藏IEEE2026-04-17 收录
下载链接：
https://ieee-dataport.org/documents/mim-vqa
下载链接
链接失效反馈官方服务：
资源简介：
The MIM-VQA dataset is designed to closely mirror real clinical scenarios by anchoring preset intents with high precision and incorporating samples that contain multiple interwoven intents. This structure allows for a more rigorous evaluation of multi-intent processing capabilities in AI frameworks. Specifically, MIM-VQA was created using clinician-curated templates and generated via GPT-4, followed by strict filtering based on faithfulness and relevance metrics. The dataset enables comprehensive assessment of compositional reasoning and intent-aware natural language generation in medical visual question answering tasks.
提供机构：
Keke Dou
