Information
中文 AI for Science AI for Science Evaluation System Advancing the Frontiers of Science and AI View Leaderboards AI4Science"Data + LLM" Assessment Framework AI4Science Assessing LLM Capabilities in Scientific Domains Recommendation of AI-Ready Scientific Data AI-Ready Links Organization Email: scihorizon@cnic.cn All Rights Reserved. Beijing ICP NO.09112257-158 The SciHorizon platform, developed by the Computer Network Information Center of the Chinese Academy of Sciences, is the world's first dedicated assessment platform for AI for Science (AI4Science). It establishes a comprehensive assessment framework designed to systematically benchmark AI4Science readiness from both large language models (LLMs) and scientific data perspectives. This framework is designed to assess the performance of AI-driven large language models (LLMs) across Mathematics, Physics, Chemistry, Life Sciences, and Earth & Space Sciences. The assessment is structured around five core indicators: Knowledge, Understanding, Reasoning, Multimodality, and Values. These indicators are further refined into 16 assessment dimensions, encompassing aspects such as knowledge authenticity, scientific fact comprehension, numerical reasoning, scientific chart interpretation, and adherence to academic integrity. Representative open-source and closed-source LLMs, sourced from both domestic and international contexts, undergo a rigorous evaluation based on this structured framework. For the intelligent application of scientific data, high-impact datasets released in recent years in the fields of earth sciences and life sciences are selected, we introduce a generalizable framework for assessing AI ready scientific data, encompassing four key dimensions—Quality, FAIRness, Explainability, and Compliance—which are subdivided into 15 sub-dimensions. Under the premise of ensuring high data quality, the semantic richness of the data and the machine-actionable capability are strengthened, and application scenarios are given as recommendations.