#TL;DR（中文）

• code复制

  Science Q&A
  的关键不是“背知识”，而是：能否从给定
  code复制

  context
  中做正确抽取与定位。
• 适合用来测试：阅读理解、信息抽取、以及在不确定时输出 “Unsure about answer”。
• 落地时建议：把
  code复制

  context
  切片、限制输出长度、并加
  code复制

  evidence
  字段或引用片段。

#Background

The following prompt tests an LLM's capabilities to perform science question answering.

#How to Apply（中文）

这个 prompt 是一个典型的 “answer based on context” 模式：

• 把你的业务资料（FAQ、policy、product spec、research notes）作为
  code复制

  Context
• 把用户问题作为
  code复制

  Question
• 让模型只基于
  code复制

  Context
  回答，缺信息就输出 “Unsure about answer”

#How to Iterate（中文）

1. 输出格式结构化：要求
   code复制

   Answer
   +
   code复制

   Evidence
   （从
   code复制

   Context
   复制关键句）
2. 控制回答长度：短回答更容易对齐输入证据
3. 加
   code复制

   abstain
   规则：不确定就 “Unsure about answer”，不要猜
4. 多问题批处理：对同一
   code复制

   Context
   一次问多个
   code复制

   Question
   ，用
   code复制

   evaluation
   比较一致性

#Self-check rubric（中文）

• code复制

  Answer
  是否能在
  code复制

  Context
  中找到对应证据？
• 是否出现
  code复制

  Context
  之外的信息（潜在
  code复制

  hallucination
  ）？
• 是否正确执行 abstain 规则（不确定就 “Unsure about answer”）？

#Practice（中文）

练习：用一段 200-400 字的内部文档当

，写 5 个问题覆盖：

• 事实抽取（who/what/when）
• 因果/目的（why）
• 数值/条件（how many/under what condition）
• 反事实（如果缺证据，必须 “Unsure about answer”）

#Prompt

markdown复制
Answer the question based on the context below. Keep the answer short and concise. Respond "Unsure about answer" if not sure about the answer.

Context: Teplizumab traces its roots to a New Jersey drug company called Ortho Pharmaceutical. There, scientists generated an early version of the antibody, dubbed OKT3. Originally sourced from mice, the molecule was able to bind to the surface of T cells and limit their cell-killing potential. In 1986, it was approved to help prevent organ rejection after kidney transplants, making it the first therapeutic antibody allowed for human use.

Question: What was OKT3 originally sourced from?
Answer:

#Code / API

#OpenAI (Python)

python复制
from openai import OpenAI

client = OpenAI()

response = client.chat.completions.create(
    model="gpt-4",
    messages=[
        {
            "role": "user",
            "content": "Answer the question based on the context below. Keep the answer short and concise. Respond \"Unsure about answer\" if not sure about the answer.\n\nContext: Teplizumab traces its roots to a New Jersey drug company called Ortho Pharmaceutical. There, scientists generated an early version of the antibody, dubbed OKT3. Originally sourced from mice, the molecule was able to bind to the surface of T cells and limit their cell-killing potential. In 1986, it was approved to help prevent organ rejection after kidney transplants, making it the first therapeutic antibody allowed for human use.\n\nQuestion: What was OKT3 originally sourced from?\nAnswer:",
        }
    ],
    temperature=1,
    max_tokens=250,
    top_p=1,
    frequency_penalty=0,
    presence_penalty=0,
)

#Fireworks (Python)

python复制
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": "Answer the question based on the context below. Keep the answer short and concise. Respond \"Unsure about answer\" if not sure about the answer.\n\nContext: Teplizumab traces its roots to a New Jersey drug company called Ortho Pharmaceutical. There, scientists generated an early version of the antibody, dubbed OKT3. Originally sourced from mice, the molecule was able to bind to the surface of T cells and limit their cell-killing potential. In 1986, it was approved to help prevent organ rejection after kidney transplants, making it the first therapeutic antibody allowed for human use.\n\nQuestion: What was OKT3 originally sourced from?\nAnswer:",
        }
    ],
    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,
)