CodeEvolve: An open source evolutionary coding agent for algorithm discovery and optimization

2025-10-15

AI Summary

This work addresses the challenging computational problem of automated algorithm discovery and optimization. We propose an LLM-driven evolutionary programming agent framework featuring three core innovations: (1) a context-window-based “inspiration crossover” mechanism that leverages LLMs to generate semantically aware program fragment recombinations; (2) an island-model population structure that enforces evolutionary isolation to preserve diversity and mitigate premature convergence; and (3) a dynamic meta-prompting strategy that adaptively refines LLM prompt templates to broaden the search space. Evaluated on multiple mathematical reasoning and program synthesis benchmarks, our framework significantly outperforms AlphaEvolve. The complete system is open-sourced, establishing a reproducible and scalable paradigm for algorithm automation through synergistic integration of large language models and evolutionary computation.

Technology Category

• Search and Optimization: Evolutionary Computation
• Machine Learning: Evolutionary Learning
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

In this work, we introduce CodeEvolve, an open-source evolutionary coding agent that unites Large Language Models (LLMs) with genetic algorithms to solve complex computational problems. Our framework adapts powerful evolutionary concepts to the LLM domain, building upon recent methods for generalized scientific discovery. CodeEvolve employs an island-based genetic algorithm to maintain population diversity and increase throughput, introduces a novel inspiration-based crossover mechanism that leverages the LLMs context window to combine features from successful solutions, and implements meta-prompting strategies for dynamic exploration of the solution space. We conduct a rigorous evaluation of CodeEvolve on a subset of the mathematical benchmarks used to evaluate Google DeepMind's closed-source AlphaEvolve. Our findings show that our method surpasses AlphaEvolve's performance on several challenging problems. To foster collaboration and accelerate progress, we release our complete framework as an open-source repository.

Problem

• Combining LLMs with genetic algorithms for computational problems
• Maintaining population diversity through island-based genetic algorithms
• Surpassing AlphaEvolve performance on mathematical benchmark problems

Innovation

• Combines LLMs with genetic algorithms
• Uses island-based algorithm for diversity
• Implements inspiration-based crossover mechanism