LLM Framework for Queueing Simulation Translation
A novel framework utilizes large language models (LLMs) to convert descriptions of conceptual queueing systems into functional SimPy scripts. The research introduces a category-template methodology aimed at ensuring comprehensive mechanism coverage, alongside a phased adaptation process that focuses on structured event logic and prevalent simulation errors. When tested on reserved tasks, the adapted models demonstrated enhancements in executability, adherence to output formats, and consistency between instructions and mechanisms across various queueing scenarios, including basic, behavioral, and networked systems, thereby increasing the reliability of the generated scripts.
Key facts
- The framework targets SimPy-based queueing model translation.
- It uses a category-template framework for mechanism coverage.
- A staged adaptation workflow addresses structured event logic and simulation-specific failure modes.
- Testing on held-out tasks showed improvements in executability, output-format compliance, and instruction-mechanism consistency.
- The approach covers basic, behavioral, and networked queueing settings.
- The goal is to reduce manual effort in verifying queueing logic.
- LLMs alone often produce executable scripts with incorrect arrival, routing, interruption, or reporting logic.
- The study is published on arXiv with ID 2601.06543.
Entities
Institutions
- arXiv