Plasma GraphRAG: Physics-Grounded Parameter Selection for Gyrokinetic Simulations

2026-04-07

AI Summary

This study addresses the inefficiency and inconsistency in parameter selection for gyrokinetic plasma simulations, which traditionally rely on manual literature review. To overcome this limitation, the authors introduce Graph Retrieval-Augmented Generation (GraphRAG) into the field for the first time, integrating large language models with a domain-specific knowledge graph constructed from plasma physics literature. Operating under physical constraints, the proposed framework enables automated recommendation of parameter ranges. By leveraging structured domain knowledge, the method enhances the accuracy and scientific credibility of recommendations while significantly mitigating model hallucinations. Experimental results demonstrate an overall improvement of over 10% across five key metrics—comprehensiveness, diversity, traceability, hallucination rate, and empowerment—with the hallucination rate reduced by up to 25%.

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 scientific discovery across complex, data-rich domains.

Problem

• gyrokinetic simulations
• parameter selection
• plasma physics
• scientific discovery
• knowledge graph

Innovation

• GraphRAG
• gyrokinetic simulation
• knowledge graph
• large language models
• physics-grounded parameter selection