3 ms·
Prompt Repetition Improves Non-Reasoning LLMs
- ajoy 7mo ago"When not using reasoning, repeating the input prompt improves performance for popular models (Gemini, GPT, Claude, and Deepseek) without increasing the number of generated tokens or latency."