The Search_2 dataset consists of (1) a set of 44 high-resolution digital color images of different complex natural scenes, (2) the ground truth corresponding to each of these scenes,
HR-MMSearch是一个基准测试数据集，旨在评估多模态大语言模型在复杂视觉任务中的代理推理和搜索能力。该数据集由商汤科技研究团队在论文《SenseNova-MARS: Empowering Multimodal Agentic Reasoning and Search via Reinforcement Learning》中提出。数据集特点包括高分辨率图像输入、知识密集型问题、搜索驱动和多领域
The Search_2 dataset consists of (1) a set of 44 high-resolution digital color images of different complex natural scenes, (2) the ground truth corresponding to each of these scenes,
