跳到主要内容Skip to main content

提示词Prompt

Prompt

  • 模型能力Model capabilities
  • 第一站 · 进门Station 1 · Getting In
  • 引用核验 2026-11-09refs checked 2026-11-09
解决什么问题What problem it solves

你给模型的指令。说得越具体,模型的输出越准。The instructions you give the model, and often the main lever you actually control.

这是什么What it is

输入给模型的那段文字。模型只根据你写的来推断意图。

同一个模型,提示词不同,效果差距可以非常大。

The input text shaping the model's behaviour. In a chat interface it is what you type; in an API it is part of the request.

Role, task, context, constraints and output format: miss one and the model fills the gap with a default assumption.
角色、任务、上下文、约束和输出格式,五件事缺一件,模型就会用默认假设补空。Role, task, context, constraints and output format: miss one and the model fills the gap with a default assumption.
为什么需要知道Why it matters

这是投入产出比最高的一个词——不用换模型,光改提示词就能明显变好。For using models, prompt quality often matters more than model choice — the same model behaves very differently under different phrasing.

别名Chinese aliases

Prompt

出处Sources
  • OpenAI 提示工程指南OpenAI Prompt engineering guide ↗
  • Anthropic 提示工程总览Anthropic Prompt engineering overview ↗