flâneur

Prompt Repetition Improves Non-Reasoning LLMs

arxiv.org · 3,021 words · saved by 1 readers

This is experimental HTML to improve accessibility. We invite you to report rendering errors. Use Alt+Y to toggle on accessible reporting links and Alt+Shift+Y to toggle off. Learn more about this project and help improve conversions. 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. LLMs are often trained as causal language models, i.e. past tokens cannot attend to future tokens. Therefore, the order of the tokens in a user’s query can affect prediction performance. For example, a query of the form “ ” often performs differently from a query of the form “ ” (see options-first vs. question-first in Figure 1). We propose to repeat the prompt, i.e. transform the input from “ ” to “ ”. This enables each prompt token to attend to every other prompt token, addressing the above. When not using re

Prompt Repetition Improves Non-Reasoning LLMs Yaniv Leviathan Google Research leviathan@google.com Matan Kalman 1 1 footnotemark: 1 Google Research matank@google.com Yossi Matias Google Research yossi@google.com Equal contribution. 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. 1 Prompt Repetition LLMs are often trained as causal language models, i.e. past tokens cannot attend to future tokens. Therefore, the order of the tokens in a user’s query can a

related reading