ImprovEvolve: Basin-Hopping Meets LLM-Guided Evolutionary Search
An AlphaEvolve-inspired algorithm that decomposes LLM-guided evolutionary search into initialization, local improvement, and perturbation operators.
How well do evolved initialization, improvement, and perturbation operators transfer beyond the optimization problems on which they were discovered?
ImprovEvolve is a useful refinement of the AlphaEvolve pattern. Instead of asking an LLM to evolve one monolithic optimizer, it evolves specialized operators for initialization, local improvement, and perturbation.
Results to watch
The paper reports new state-of-the-art packings for multiple hexagon-in-hexagon cases, a stronger lower bound for the second autocorrelation inequality, and improvements for many spherical-code instances.
ASI relevance
The big lesson is cognitive load management. If we want LLM-guided search to scale, we may need decomposed operator libraries rather than heroic single-shot program synthesis.