Vireo
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/steven-ccq/VisualReasoner
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个通过最少到最多合成方法构建的综合视觉推理数据集，包含了5万个视觉推理示例。该数据集旨在以即插即用的方式提升各种视觉-语言模型的推理能力，并且已经通过人工检查对其质量进行了评估。数据集的规模为5万个示例，任务类型为视觉推理。
This is a comprehensive visual reasoning dataset constructed via the least-to-most synthesis method, containing 50,000 visual reasoning instances. It aims to improve the reasoning capabilities of various vision-language models in a plug-and-play manner, and its quality has been evaluated through manual inspection. The dataset focuses on visual reasoning tasks and includes 50,000 such samples.
提供机构：
Authors of the paper
