llm-optimization

Google researchers found that simply repeating your prompt—copying and pasting it twice—dramatically improves LLM accuracy on non-reasoning tasks, with gains up to 76% and zero performance degradation.