sensenova/HR-MMSearch

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、搜索驱动和多领域覆盖（如体育、娱乐与文化、科学与技术、商业与金融、游戏、学术研究、地理与旅行等）。每个样本包含样本ID、查询文本、查询图像路径、真实答案、难度级别和类别等信息。

HR-MMSearch is a benchmark designed to evaluate the Agentic Reasoning and Search capabilities of Multimodal Large Language Models in complex visual tasks. This dataset was introduced by SenseTime Research in the paper *SenseNova-MARS: Empowering Multimodal Agentic Reasoning and Search via Reinforcement Learning*. Key features include High-Resolution Images, Knowledge-Intensive questions, Search-Driven tasks, and Multi-Domain Coverage (e.g., Sports, Entertainment & Culture, Science & Technology, Business & Finance, Games, Academic Research, Geography & Travel, etc.). Each sample contains fields such as sample_id, query, query_image, ground_truth, difficulty, and category.