AI & LLMs · Intermediate
幻觉(Hallucination)
模型一本正经地说出并不存在的事实——虚构的书名、打不开的链接、记错的日期。它语气越笃定,越容易让人信以为真。
Some fields on this page have no English version yet — they are shown in the original Chinese.
TL;DRIn 30 seconds
- 模型一本正经地说出并不存在的东西——虚构的书名、打不开的链接、记错的日期。
- 它说这些时并不含糊,反而格外肯定,人很容易被这种语气带过去。
- 根源是它在往下续写,不是在核实,所以关键字段仍要人工对一遍。
OverviewIn detail
幻觉指模型生成听起来合理、实际上不成立的内容。虚构的文献、编造的引用、张冠李戴的日期、并不存在的产品型号,都可能冒出来。麻烦的是它说这些时并不含糊,反而格外肯定,人很容易被这种语气带过去。它也不会标出哪句有把握、哪句是猜的。
根源在于模型的机制是续写,不是查证。它擅长判断下一句怎么写最通顺,却不具备“知道自己不知道”的能力。碰到没把握的地方,它仍会顺着一句话的势头往下编,还编得有模有样。训练的目标是让回答更像真的,而不是让回答更接近真相,所以这不是它笨,而是它的工作方式本身如此。
最常见的触发场合是问细节。让它列一份资料来源,它可能给出一个格式规范、实际却查不到的出处;让它报一个具体数字,它可能给出一个看上去合理却错的值。越是被问到冷门细节,它越容易顺着语气往下补。给它材料、让它基于材料回答,能明显减少这类编造,但减少不等于消除。
因此有一条实用底线:凡是数字、引用、人名、法条、链接这类错了要出事的字段,都自己核一遍。把关键结论先标出来再逐条核实,比通读一遍更靠得住。也别把模型给出的出处直接当证据转手引用,那等于把编造又传播了一遍。
MythsCommon misunderstandings
- ✕以为它是故意骗你 ✓它压根分不清哪句真哪句是顺着语气编的,没有撒谎这个念头
- ✕以为语气肯定就更可信 ✓语气和真假完全无关,越是冷门细节越容易编得有模有样
- ✕以为给了资料就彻底不编了 ✓给材料能明显减少编造,但减少不等于消除,仍要抽查
In practiceWhere you will meet it
- 让它列一份资料清单,逐条去查那个出处是不是真的存在
- 把方案里的数字、人名、日期单拎出来,自己核一遍再交出去
- 看到一段漂亮引用时,先搜原文,别直接转手发出去
SummaryThe key points
- 模型会生成看似合理但实际错误的内容
- 它说错话时往往语气格外肯定
- 根源是它的机制在续写而非核实
- 数字、引用、出处必须人工复核
FAQ
What does 幻觉(Hallucination) mean?
模型一本正经地说出并不存在的东西——虚构的书名、打不开的链接、记错的日期。它说这些时并不含糊,反而格外肯定,人很容易被这种语气带过去。根源是它在往下续写,不是在核实,所以关键字段仍要人工对一遍。
Where will I run into 幻觉(Hallucination)?
让它列一份资料清单,逐条去查那个出处是不是真的存在;把方案里的数字、人名、日期单拎出来,自己核一遍再交出去;看到一段漂亮引用时,先搜原文,别直接转手发出去。
What do people most often get wrong about 幻觉(Hallucination)?
以为它是故意骗你 — but 它压根分不清哪句真哪句是顺着语气编的,没有撒谎这个念头;以为语气肯定就更可信 — but 语气和真假完全无关,越是冷门细节越容易编得有模有样。
What are the key points about 幻觉(Hallucination)?
模型会生成看似合理但实际错误的内容;它说错话时往往语气格外肯定;根源是它的机制在续写而非核实;数字、引用、出处必须人工复核。
Common pitfalls 模型给出的文献出处和链接看起来再规范,也要自己查一遍,直接引用很容易把错传出去。
Entry 30 of 186 · All entries
Keep readingRelated reading
AI Agent
An AI program that sets its own goal, breaks it into steps and calls tools to get the job done. You hand it a task and it works out how to do it — instead of only answering one question at a time.
LLM
A very large program that learned to predict the next piece of text after reading an ocean of it. Whatever you type, it is guessing what is most likely to come next — writing, translating and answering questions are all the same thing used differently.
嵌入(Embedding)
把一段文字换算成一串数字,让电脑能比较哪两句话意思更接近。搜索和知识库能在换了说法时也找对答案,靠的就是它。
Claude
A large language model built by Anthropic, known for understanding long documents, writing well and writing code. It comes both as a chat assistant on the web and phone, and in a dedicated form for programming.
Prompt
The message you send to an AI — the instruction or question you give it. How well that passage is written decides whether the answer that comes back is on target.
MCP
A standard interface that lets AI connect to outside tools and data in one uniform way. Build a tool to it once and any AI assistant can call it, saving you from writing a separate integration for each tool.
Where to go nextExplore more
Back to the skills themselves
Most of the concepts covered here come from skills that install into your AI client with a single command.