ReVEL: Multi-Turn Reflective LLM-Guided Heuristic Evolution via Structured Performance Feedback

“ReVEL leverages multi-turn LLM reasoning to automatically design better heuristics for NP-hard optimization problems, moving beyond one-shot code generation. This hybrid approach could significantly reduce the expertise required to tackle computationally complex real-world challenges across logistics, scheduling, and design optimization.”