
Verifying AI-Generated Code
Formal Verification Techniques for LLM Output
Marmaragan is a novel tool that enables formal verification of code produced by large language models through automated generation of SPARK annotations for Ada programs.
- Addresses the critical reliability gap in LLM-generated code through formal verification
- Automates the creation of SPARK annotations, making verification more accessible
- Benchmarks the effectiveness of verification against existing programs
- Bridges advanced AI capabilities with established software engineering practices
This research provides a practical solution for engineering teams looking to leverage LLM code generation while maintaining high reliability standards for critical systems.
Original Paper: Verifying LLM-Generated Code in the Context of Software Verification with Ada/SPARK