## Prompt Repetition: A Simple Technique to Improve LLMs A recent study has revealed that repeating prompts can lead to notable improvements in the performance of large language models (LLMs), without negatively impacting latency. The technique applies in scenarios where reasoning is not a crucial factor. The research indicates that prompt repetition does not alter the length or format of the generated outputs. The results suggest that this strategy could be adopted as a default setting for many models and tasks, especially when reasoning is not required. Researchers found significant gains on several benchmarks, with Deepseek as the only open-source model evaluated. The study opens new perspectives on how to optimize LLM performance through prompt engineering techniques.