[2506.19863] Exploring the Capabilities of the Frontier Large Language Models for Nuclear Energy Research

1 2 minutes read

250619863 Exploring the Capabilities of the Frontier Large Language Models.png

[Submitted on 10 Jun 2025]

Authors:Ahmed Al -Miladin, Mohamed Angogar, Rick Archibald, Tom Beck, Erban Biswas, Rake Bostelman, Weiss Burwer, Chris Brian, Christopher Cali, Sihanger Celik, Rajni Shal, Jongk Choi, Arnam Chaudhry, Mark Sianchius, Pascuale, Lisa Fassino, Ana Ginaru, Yashika Ghai, Luke Gibson, Qian Gong, Christopher Greulich, Scott Greenwood, Cory Hauck, Ehab Hassan, Rinkle Junja, Soyoung Kang, Scott Klasky, Atul Kumar, Vineet Kumar , McConnell, Forecan Oz, Anant Raj, Pradib Ramohali, Marie Romidin, Samantha Sabatino, Jose Salsido Perez, Nathan de, Erban Cerable, Bunam Shukr, Tim Uninken, Xiao Feng, Praishant Jin, Tom Evans, Brasana, Brasan.

View the PDF file for the paper entitled Explore the capabilities of the large border linguistic models for nuclear energy research, written by Ahmed Aldin and 48 other books

PDF HTML (experimental) view

a summary:Artificial Intelligence for the Nuclear Energy Workshop at the OAK RIDGE National Laboratory has evaluated the capabilities of large language models (LLMS) to accelerate fusion and fission research. Fourteen multidisciplinary teams explore various nuclear science challenges using ChatGPT, Gemini, Claude, and other Amnesty International models over one day. Applications ranged from the development of basic models to control the fusion reactor to the Monte Carlo simulation automation, predict the deterioration of materials, and the design of experimental programs for advanced reactors. The difference was used in an organized workflow that combines immediate engineering, deep search capabilities, frequency refining to generate hypotheses, the primary model code, and research strategies. The main results show that LLMS excels in exploration in the early stage, the synthesis of literature, the design of workflow, and the success of the search gaps successfully and the generation of reasonable experimental frameworks. However, large restrictions, including difficulties in new material designs, generate advanced code of modeling and simulation, and the details of the field that require verification of experts’ health. Successful results from expert -based engineering and artificial intelligence treatment as a complementary tool instead of replacing physics -based methods. The workshop checks the capabilities of artificial intelligence to accelerate nuclear energy research through rapid repetition and synthesis through disciplines, highlighting the need for nuclear attributed data groups, automation of workflow, and developing a specialized model. These results provide a road map to integrate artificial intelligence tools into the functioning of nuclear science, which may reduce the development cycles of the safer and more efficient nuclear energy systems while maintaining strict scientific standards.