Atomic Canyon 与爱达荷国家实验室推出 NuclearBench 核能大模型开源基准
科技公司 Atomic Canyon 与美国能源部爱达荷国家实验室（INL）11月5日推出 NuclearBench，这是首个专门评测大语言模型核能知识的开源基准工具，在 FermiBench 的基础上扩展而成。
NuclearBench 围绕约280道核能专业问题构建，从准确性与忠实性（faithfulness）两个维度为模型的回答打分，旨在为核工业引入超级智能（SI）工具提供可核查的评测标准。双方表示，该基准即日起开源可用。
（来源：GlobeNewswire）
Atomic Canyon and Idaho National Lab launch NuclearBench, an open benchmark for nuclear AI models
Atomic Canyon and the U.S. Department of Energy's Idaho National Laboratory (INL) launched NuclearBench on November 5, the first open-source benchmark designed to test large language models on nuclear questions. It extends the earlier FermiBench.
Built around roughly 280 nuclear questions, NuclearBench scores model answers on both accuracy and faithfulness, aiming to give the nuclear industry a verifiable yardstick as it adopts super-intelligence (SI) tools. The benchmark is available as open source.
(Source: GlobeNewswire)
来源：GlobeNewswire
发布：GSM·美国（发布链接）
图片来源：爱达荷国家实验室设施（图：Wikimedia Commons，CC BY-SA 4.0） / An Idaho National Laboratory facility (Wikimedia Commons, CC BY-SA 4.0)
GSM Global · 本篇英文版 English edition of this story
