InternScience/SGI-WetExperiment

SGI-Bench数据集旨在通过科学家对齐的工作流程评估大型语言模型（LLMs）的科学通用智能（SGI）。该数据集涵盖10个学科，包含1000多个由专家精心策划的样本，灵感来源于《科学》杂志的125个重大问题。数据集包括四个主要任务家族：科学深度研究、创意生成、干/湿实验和多模态实验推理。数据集的结构旨在评估科学研究的完整探究周期，包括审议、构思、行动和感知。README还详细介绍了数据集的构建、评估框架以及使用数据集的快速入门指南。

The SGI-Bench dataset is designed to evaluate the Scientific General Intelligence (SGI) of large language models (LLMs) through scientist-aligned workflows. The dataset spans 10 disciplines and includes over 1,000 expert-curated samples inspired by Sciences 125 Big Questions. It covers four main task families: scientific deep research, idea generation, dry/wet experiments, and multimodal experimental reasoning. The dataset is structured to assess the full inquiry cycle of scientific research, including deliberation, conception, action, and perception. The README also details the datasets construction, evaluation framework, and quick start instructions for using the dataset.