Daily Productive Sharing 690 - Prompt Engineering

One helpful tip per day:)

Lilian Weng reviewed the prompts used by large language models (LLMs) in recent years and provided detailed explanations:

  1. Zero shot prompt: Directly giving instructions -- this is the most basic instruction.
  2. Few shot prompt: Providing example(s) and giving instructions -- this is more effective than the previous one, but it consumes more tokens.
  3. Chain of thoughts (CoT): Providing certain reasoning steps or reasoning instructions -- this is the most effective for complex problems in larger LLMs.

If you enjoy today's sharing, why not subscribe

Need a superb CV, please try our CV Consultation


Lilian Weng 回顾了近几年大语言模型 (LLM) 所使用的 prompts,给出了详细的介绍:

  1. zero shot prompt:直接给出指令 -- 这是最基本的指令
  2. few shot prompt: 给出示例e,并给出指令 -- 效果比前面这种好,但是会消耗更多的 tokens
  3. chain of thoughts (CoT): 给出一定的推理步骤,或者推理指令 -- 针对复杂问题,在较大的 LLM 下,这种效果最好。

如果你喜欢我们的内容,不如支持我们 :)

需要更棒的简历,不妨试试我们的 CV Consultation