Reproduce “LLM4PLC: Harnessing Large Language Models for Verifiable Programming of PLCs in Industrial Control Systems” 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
Mohamad Fakih, Rahul Dharmaji, Yasamin Moghaddas, Gustavo Quiros Araya, Oluwatosin Ogundare, Mohammad Abdullah Al Faruque
The LLM4PLC system automates the verifiable programming of Programmable Logic Controllers (PLCs) in industrial control systems by integrating Large Language Models with external verification tools and an iterative feedback loop. This approach significantly improves the pass rate of generated Structured Text (ST) code, achieving up to 72.5% compilation success and higher expert-rated code quality compared to unaugmented LLMs.
Key insights
- Integrating external automated verification tools into an iterative feedback loop with LLMs substantially improves the verifiability and correctness of generated code for specialized domains.
- Domain-specific fine-tuning using PEFT (LoRAs) enables smaller, specialized LLMs (like Code Llama 34B) to surpass the performance of larger, general-purpose LLMs (like GPT-4) in generating niche programming languages.
- A Model-Based Design approach, beginning with LLM-generated FSM schematics, provides a structured and predictable foundation for PLC code generation, enhancing debuggability and alignment with operational requirements.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.