CHECKMATE Evolves Code to Solve Industrial Optimization Problems

A new tool called CHECKMATE demonstrates that algorithm generation via code evolution can solve combinatorial and optimization problems without requiring experts to specify solution derivation methods. The system relies solely on formal specifications to ensure correctness and enable performance evaluation, while natural language descriptions guide the evolutionary process. Tested on configuration and scheduling problems from two industrial domains, the evolved algorithms consistently outperformed existing state-of-the-art solvers. This paradigm shift eliminates the need for specialized solvers and expert-designed heuristics, allowing experts to focus on defining what solutions are rather than how they are derived. The research is detailed in arXiv:2605.31049.