SeaEvo: Advancing Algorithm Discovery with Strategy Space Evolution

TL;DR

SeaEvo introduces a modular strategy-space layer that enhances LLM-guided evolutionary search by organizing strategic reasoning as persistent population-level state, leading to significant performance improvements.

Contribution

It presents a novel strategy-space layer that converts language-level reasoning into structured population state, improving search effectiveness without altering core algorithms.

Findings

Achieved a 20.6% average relative improvement across four benchmarks.

The best run on Prism scored 3 times higher than baseline.

Enhanced the ability to preserve promising strategies and avoid saturation.

Abstract

Large Language Model (LLM)-guided evolutionary search is increasingly used for automated algorithm discovery, yet most current methods track search progress primarily through executable programs and scalar fitness. Even when natural-language reasoning is used through heuristic descriptions or reflection, it typically remains transient mutation context or unstructured memory, rather than organized as persistent population-level state over strategic directions. As a result, evolutionary search can struggle to distinguish syntactically different implementations of the same idea, preserve lower-fitness but strategically promising directions, or detect when an entire family of strategies has saturated. We introduce \model, a modular strategy-space layer that turns language-level strategic reasoning into first-class population-level evolutionary state in LLM-driven program search. \model represents each candidate program with an explicit natural-language strategy, clusters the archive by strategy semantics, retrieves behaviorally complementary inspirations, and periodically navigates the strategy landscape to avoid saturated directions. Without modifying the underlying evolutionary algorithms, \model improves existing evolutionary backbones across algorithm discovery, systems optimization, and agent-scaffold design tasks in most settings. Across four systems benchmarks, \model achieves a 20.6% average relative improvement, with the best single run on Prism scoring 3$\times$ higher. These results suggest that persistent strategy representations provide a practical mechanism for improving the effectiveness and cost-efficiency of LLM-guided evolutionary search, pointing toward compound AI systems whose search capabilities benefit from the structured accumulation and reuse of algorithmic strategies.