提示工程
定义
提示工程是设计输入文本（提示）以从 LLM 获得期望行为的实践: task format, few-shot examples, chain-of-thought, role-playing, and constraints.
它是 the primary way to steer LLMs without fine-tuning: you control context, format, and examples in the prompt. Combined with RAG, prompts often include retrieved passages; with agents, they define tool use and 推理 style.
工作原理
你编写一个提示（系统消息、任务描述、约束）并可选地添加示例（few-shot）。LLM takes this as input and produces an output. Zero-shot uses only instructions; few-shot adds example input-output pairs so the model infers the task. Chain-of-thought (see CoT) asks the model to “think 逐步” to improve 推理. Structured output (例如 “respond in JSON”) can be enforced via parsing or API options. Iterate on prompt wording and examples, and evaluate on a dev set to improve reliability.
应用场景
Prompt engineering matters whenever you call an LLM: it shapes behavior, format, and 推理 without changing weights.
- Steering chat and task completion (role, format, examples)
- Eliciting 推理 (chain-of-thought) for math or logic
- Constraining outputs (JSON, length, tone) for APIs or UX