The best AI optimizers succeed by acting as narrow refiners rather than creative explorers.
April 24, 2026
Original Paper
What Makes an LLM a Good Optimizer? A Trajectory Analysis of LLM-Guided Evolutionary Search
arXiv · 2604.19440
The Takeaway
Standard intuition suggests that AI solves hard problems through creative leaps. This analysis of evolutionary search shows that the most successful models are those that aggressively restrict their focus. Weaker models often suffer from semantic drift because they wander too far from the core problem. The winning strategy for an LLM is to act as a local refiner that polishes existing solutions. This suggests that the value of AI in optimization lies in its discipline rather than its imagination.
From the abstract
Recent work has demonstrated the promise of orchestrating large language models (LLMs) within evolutionary and agentic optimization systems. However, the mechanisms driving these optimization gains remain poorly understood. In this work, we present a large-scale study of LLM-guided evolutionary search, collecting optimization trajectories for 15 LLMs across 8 tasks. Although zero-shot problem-solving ability correlates with final optimization outcomes, it explains only part of the variance: mode