Evolutionary thoughts: integration of large language models and evolutionary algorithms

TL;DR

This work tackles the challenge of expensive exploration in evolutionary algorithms for program synthesis by marrying large language models with a fast, scalable evaluation framework. The proposed hybrid approach uses LLM-guided seed generation and mutation, coupled with an efficient C++/GPU evaluation backend, to focus search and accelerate fitness assessments. Empirical results on synthetic induction tasks (Count, Max-Min, Inverse, Sort) show that LLM guidance improves accuracy and reduces program length, and that ensembling multiple runs yields robust, often perfect, performance on some tasks. The findings suggest that semantic priors from LLMs can meaningfully augment evolutionary search, enabling scalable automatic programming with improved reliability and efficiency.

Abstract

Large Language Models (LLMs) have unveiled remarkable capabilities in understanding and generating both natural language and code, but LLM reasoning is prone to hallucination and struggle with complex, novel scenarios, often getting stuck on partial or incorrect solutions. However, the inherent ability of Evolutionary Algorithms (EAs) to explore extensive and complex search spaces makes them particularly effective in scenarios where traditional optimization methodologies may falter. However, EAs explore a vast search space when applied to complex problems. To address the computational bottleneck of evaluating large populations, particularly crucial for complex evolutionary tasks, we introduce a highly efficient evaluation framework. This implementation maintains compatibility with existing primitive definitions, ensuring the generation of valid individuals. Using LLMs, we propose an enhanced evolutionary search strategy that enables a more focused exploration of expansive solution spaces. LLMs facilitate the generation of superior candidate solutions, as evidenced by empirical results demonstrating their efficacy in producing improved outcomes.