发表机构
Thomas Jefferson National Accelerator Facility（托马斯·杰斐逊国家加速器设施）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究以杰斐逊实验室GlueX实验为试点，提出数据需先满足FAIR原则方可AI就绪，并构建了基于确定性指标的AI就绪评估框架，通过评估数据生态系统与现有验证软件，给出配套建议。
AI 中文摘要
美国能源部的项目日益要求实验数据不仅被保存，而且能被人工智能（AI）系统使用，但目前尚无公认的标准来定义数据集何时达到AI就绪。我们认为，数据必须先满足FAIR（可查找、可访问、可互操作、可重用）原则，然后才能达到AI就绪，且AI就绪是一个需要独立评估的独特属性。以杰斐逊实验室D厅的GlueX实验为试点，我们绘制了数据生态系统图，依据FAIR原则对其进行评估，并定义了一个AI就绪框架，其标准由确定性指标支撑。我们还评估了现有的验证软件，以区分哪些可以用现成工具评估，哪些需要领域特定的开发。评估产生了一系列发现，每项发现都配有一条建议，其中一些已投入生产或已原型化，另一些则计划作为后续步骤。
英文摘要
Department of Energy programs increasingly require that experimental data be not only preserved but usable by artificial intelligence (AI) systems, yet no accepted standard defines when a dataset is AI-ready. We argue that data must be FAIR (Findable, Accessible, Interoperable, Reusable) before they can be AI-ready, and that AI readiness is a distinct property requiring its own assessment. Using the GlueX experiment in Hall D at Jefferson Lab as a pilot, we map the data ecosystem, assess it against FAIR, and define an AI-readiness framework whose criteria are backed by deterministic metrics. We also evaluate existing validation software to separate what can be assessed with off-the-shelf tools from what requires domain-specific development. The assessment yields a set of findings, each paired with a recommendation, some of which are already in production or prototyped and others planned as next steps.
Comments18 pages, 1 figure, 5 tables
