AdaptLLM/biomed-visual-instructions
收藏Hugging Face2025-03-21 更新2024-12-14 收录
下载链接:
https://hf-mirror.com/datasets/AdaptLLM/biomed-visual-instructions
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含用于后训练多模态大语言模型(MLLMs)的生物医学视觉指令任务。数据集通过视觉指令合成器生成,基于PubMedVision中的图像-描述对。数据集包括500K个图像-描述对、144K个合成的指令-响应对,以及用于单阶段领域特定后训练的500K个示例。这些数据用于训练和评估MLLMs在生物医学领域的性能。
This dataset contains biomedical visual instruction tasks for post-training multimodal large language models (MLLMs). The dataset is generated using a visual instruction synthesizer based on image-caption pairs from PubMedVision. It includes 500K image-caption pairs, 144K synthetic instruction-response pairs, and 500K examples for single-stage domain-specific post-training. These data are used to train and evaluate the performance of MLLMs in the biomedical domain.
提供机构:
AdaptLLM



