OpenEvolve: An Open Source Implementation of Google DeepMind's AlphaEvolve

OpenEvolve is an open-source implementation of Google DeepMind's AlphaEvolve, designed to harness large language models (LLMs) for evolutionary algorithm discovery. By integrating modern LLMs with evolutionary techniques, OpenEvolve enables significant advancements in code optimization and algorithm development, replicating successes such as the circle packing problem and function minimization. It operates through a flexible architecture comprising components like a prompt sampler, LLM ensemble, evaluator pool, and program database, all orchestrated to maximize throughput and evaluate candidate solutions. OpenEvolve's approach allows for the evolution of entire code files across multiple programming languages, supporting multi-objective optimization and employing a two-phase strategy to break through solution plateaus. The platform demonstrates the transformative potential of guided evolutionary search, evolving basic implementations into advanced algorithms, and invites the community to contribute to and utilize the system for diverse algorithmic challenges.