Prompt Repetition Improves Non-Reasoning LLMs [pdf] | Dark Hacker News