Reproduce “Simulation Code Generation for Fluid Systems using Large Language Models: Benchmarking Models and Prompting Strategies” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jan Marius Stürmer, Jascha Knack, Tobias Koch, Andreas Weinmann
The authors test whether an LLM can translate supplied fluid-system graphs into framework-specific simulation wrappers for WNTR and Modelica. They compare library-context strategies and agent validation. Curated code improves executability over documentation in one Modelica comparison. For Kimi-K2.5, validation raises pass@1 from 0.692 ± 0.063 to 0.923 ± 0.063, while Modelica simulation fidelity remains 0.0 ± 0.0, showing execution does not establish physical accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.