[2604.06279] Plasma GraphRAG: Physics-Grounded Parameter Selection for Gyrokinetic Simulations

Physics > Plasma Physics arXiv:2604.06279 (physics) [Submitted on 7 Apr 2026] Title:Plasma GraphRAG: Physics-Grounded Parameter Selection for Gyrokinetic Simulations Authors:Ruichen Zhang, Feda AlMuhisen, Chenguang Wan, Zhisong Qu, Kunpeng Li, Youngwoo Cho, Kyungtak Lim, Virginie Grandgirard, Xavier Garbet View a PDF of the paper titled Plasma GraphRAG: Physics-Grounded Parameter Selection for Gyrokinetic Simulations, by Ruichen Zhang and 8 other authors View PDF HTML (experimental) Abstract:Accurate parameter selection is fundamental to gyrokinetic plasma simulations, yet current practices rely heavily on manual literature reviews, leading to inefficiencies and inconsistencies. We introduce Plasma GraphRAG, a novel framework that integrates Graph Retrieval-Augmented Generation (GraphRAG) with large language models (LLMs) for automated, physics-grounded parameter range identification. By constructing a domain-specific knowledge graph from curated plasma literature and enabling structured retrieval over graph-anchored entities and relations, Plasma GraphRAG enables LLMs to generate accurate, context-aware recommendations. Extensive evaluations across five metrics, comprehensiveness, diversity, grounding, hallucination, and empowerment, demonstrate that Plasma GraphRAG outperforms vanilla RAG by over $10\%$ in overall quality and reduces hallucination rates by up to $25\%$. {Beyond enhancing simulation reliability, Plasma GraphRAG offers a methodology for accelerating scient...