核心结论：在不使用推理（reasoning）时，重复输入提示（repeatitive prompts）可以提高主流模型（如：Gemini、GPT、Claude、Deepseek）的性能表现，且不会增加生成的 tokens 数量或带来延迟。发表团队及时间：谷歌团队 2025年末具体方法：① Prompt RepetitionLLMs 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 “CONTEXT QUESTION” often performs differently from a query of the form “QUESTION CONTEXT” (see options-first vs. question-first in Figure 1). We propose to repeat the prompt, i.e. transform the input from “QUERY” to “QUERYQUERY”. Prompt Repetition Improves Non-Reasoning LLMs 解读
核心结论：在不使用推理（reasoning）时，重复输入提示（repeatitive prompts）可以提高主流模型（如：Gemini、GPT、Claude、Deepseek）的性能表现，且不会增加生成的 tokens 数量或带来延迟。发表团队及时间：谷歌团队 2025年末具体方法：① Prompt RepetitionLLMs 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 “CONTEXT QUESTION” often performs differently from a query of the form “QUESTION CONTEXT” (see options-first vs. question-first in Figure 1). We propose to repeat the prompt, i.e. transform the input from “QUERY” to “QUERYQUERY”.