 |
ML Newsnew | past | best | rss | submit | about |
| Prompt Repetition Improves Non-Reasoning LLMs [pdf] (arxiv.org) | |
| 1 point by 8ig8 227 days ago | hide | past | pdf | discuss on HN | |
|
In plain words: Simply writing the same prompt more than once before asking for an answer makes popular chat models answer better when they are not thinking step by step. It beats the usual single prompt without generating extra words or taking longer. Abstract · Prompt Repetition Improves Non-Reasoning LLMsWhen 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. | |
|
add comment on HN
Also discussed: Mar 2026 (3 points, 0 comments) · Feb 2026 (1 point, 0 comments) · Feb 2026 (1 point, 1 comment) · Feb 2026 (1 point, 0 comments) · Feb 2026 (1 point, 0 comments) · Feb 2026 (2 points, 0 comments) · Jan 2026 (2 points, 0 comments) · Jan 2026 (3 points, 0 comments) · Jan 2026 (2 points, 0 comments) · Jan 2026 (1 point, 0 comments) · Jan 2026 (2 points, 0 comments) · Dec 2025 (2 points, 1 comment)
|