Large language models can act as evolutionary optimization algorithms through a novel prompting strategy, EvoLLM, outperforming traditional baselines on various tasks.
