P
Prompt Master
Prompt Engineering 教程与提示词实战

从任务定义、示例和工作流到评测与安全边界

Prompt Engineering 教程与提示词实战Prompt Library

Adding odd numbers

math capability prompt example

本章 Prompt 决定
01要解决的 Prompt 问题

math capability prompt example

02完成后带走

一份保留任务边界的可复用 Prompt、示例集、评测记录或安全规则。

03做到什么算完成

至少跑一个代表性样本,检查结果,并记录仍需人工复核的部分。

TL;DR(中文)

  • 这是一个简单的 math reasoning 测试:先做 filtering(挑 odd numbers),再做 addition,最后判断 parity(odd/even)。
  • 适合用来观察:模型是否会按步骤执行、是否会在中间计算出错、以及是否能保持一致性。
  • 在生产里建议:要求输出每一步的中间结果(列表、求和过程、最终结论),便于 evaluation

Background

This prompt tests an LLM's mathematical capabilities by asking it to check whether the odd numbers in a group add up to an even number. It also nudges the model to break the problem into steps.

How to Apply(中文)

把这个 prompt 当成 “step-by-step 执行” 的模板即可:

  1. Identify(筛选 odd numbers)
  2. Compute(求和)
  3. Verify(判断 odd/even)

如果你要把它迁移到业务里的计算任务,核心是:把每一步写成可验收的子任务,并要求输出中间结果。

How to Iterate(中文)

  1. 结构化输出:要求输出一个小表格 numbers / is_odd / included / running_sum
  2. self-check:最后重复一遍求和,或用不同顺序再算一次
  3. 控制 temperature(数学任务通常建议更低温度以提升一致性)
  4. 对抗干扰:把列表变长、加入无效字符或更复杂格式,测试鲁棒性

Self-check rubric(中文)

  • 是否正确识别了哪些是 odd numbers?
  • 求和是否正确?是否有中间步骤错误?
  • 最终 parity 判断是否与求和一致?
  • 输出是否包含可复核的中间结果?

Practice(中文)

练习:把同样的模板应用到下面两类任务:

  • 从一串混合输入里提取满足条件的数字(例如 divisible by 3),再求和
  • 给定一串日期,筛选出周末,再统计数量

Prompt

The odd numbers in this group add up to an even number: 15, 32, 5, 13, 82, 7, 1.
Solve by breaking the problem into steps. First, identify the odd numbers, add them, and indicate whether the result is odd or even.

Code / API

OpenAI (Python)

from openai import OpenAI

client = OpenAI()

response = client.chat.completions.create(
    model="gpt-4",
    messages=[
        {
            "role": "user",
            "content": "The odd numbers in this group add up to an even number: 15, 32, 5, 13, 82, 7, 1.\nSolve by breaking the problem into steps. First, identify the odd numbers, add them, and indicate whether the result is odd or even.",
        }
    ],
    temperature=1,
    max_tokens=256,
    top_p=1,
    frequency_penalty=0,
    presence_penalty=0,
)

Fireworks (Python)

import fireworks.client

fireworks.client.api_key = "<FIREWORKS_API_KEY>"

completion = fireworks.client.ChatCompletion.create(
    model="accounts/fireworks/models/mixtral-8x7b-instruct",
    messages=[
        {
            "role": "user",
            "content": "The odd numbers in this group add up to an even number: 15, 32, 5, 13, 82, 7, 1.\nSolve by breaking the problem into steps. First, identify the odd numbers, add them, and indicate whether the result is odd or even.",
        }
    ],
    stop=["<|im_start|>", "<|im_end|>", "<|endoftext|>"],
    stream=True,
    n=1,
    top_p=1,
    top_k=40,
    presence_penalty=0,
    frequency_penalty=0,
    prompt_truncate_len=1024,
    context_length_exceeded_behavior="truncate",
    temperature=0.9,
    max_tokens=4000,
)

Reference

📚 相关资源