MedVLSynther/MedSynVQA-5K-qwen-glm
收藏Hugging Face2025-10-29 更新2025-11-01 收录
下载链接：
https://hf-mirror.com/datasets/MedVLSynther/MedSynVQA-5K-qwen-glm
下载链接
链接失效反馈官方服务：
资源简介：
MedSynVQA是一个针对医疗视觉问答任务的完整训练数据集，用于RLVR训练。它包含了不同大小的子集，如10K、5K、2K和1K样本的训练子集，以及针对不同实验设置的特定子集，例如生成器和验证器选择的消融研究、生成器选择的消融研究、验证器选择的消融研究、验证器必要性的消融研究、PMC-VQA风格提示生成的子集和为SFT训练生成的子集。
MedSynVQA is a full training dataset for medical visual question answering tasks, used for RLVR training. It includes subsets of various sizes such as 10K, 5K, 2K, and 1K training subsets, as well as specific subsets for different experimental settings, such as generator and verifier choice ablation studies, generator choice ablation studies, verifier choice ablation studies, verifier necessity ablation studies, PMC-VQA-style prompt-generated subsets, and subsets generated for SFT training.
提供机构：
MedVLSynther
