Chain-of-thought prompting

As explained in How Large Language Models Work

A prompting technique that encourages a language model to produce an intermediate sequence of reasoning steps before an answer.

Chain-of-thought prompting asks a language model to produce intermediate reasoning steps, often with “Let’s think step by step.” This can improve task performance, but it does not prove that the model thinks or understands the world; its activity is better described as calculating and generating text. section 7.3

A useful but unreliable prompt

For a difficult math problem, asking for step-by-step output may provide useful intermediate calculations. The model can still skip steps, miss calculations, or produce logically invalid reasoning. section 7.3

Figure 7.6 The context and reason why someone is wearing or doing something unusual may be in the realm of something that an LLM properly recognizes and for which it produces an appropriate response. However, it might not be possible for an LLM to reach that appropriate response without producing some intermediate text. For a math problem, this intermediate text could be useful, but the intermediate text may not always be appropriate or desirable for a user to see.

Intermediate text can help a model reach an appropriate response, yet showing that text may not always be desirable, especially outside tasks such as mathematics. section 7.3

How How Large Language Models Work covers it

How deep it goes

Explains the prompt pattern and possible mechanisms, but stops short of resolving why performance improves.

Examples

Uses “Let’s think step by step,” difficult math problems, planning, and unusual situations requiring contextual interpretation.

What you can do after

Apply it via a step-by-step prompt while checking for missing steps, calculation errors, and invalid reasoning.

How it connects

Connects chain-of-thought prompting with language models, transformers, attention mechanisms, and world models.

liveBookmanning.com

Read How Large Language Models Work on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Introduction to Generative AI, Second EditionChain-of-thought prompting is a technique that elicits intermediate reasoning steps before a model produces its final answer.
• 

  liveBookmanning.com
  Coding with AIChain-of-thought prompting is a prompting technique that guides an AI model through a problem using a sequence of intermediate reasoning steps.
• 

  liveBookmanning.com
  AI Agents in Action, Second Edition
• 

  liveBookmanning.com
  The Art of AI Product DevelopmentChain-of-thought is a prompting technique that elicits an intermediate sequence of reasoning steps before a model produces a final answer.
• 

  liveBookmanning.com
  LLMs in ProductionChain-of-thought prompting encourages a language model to produce or use intermediate reasoning steps before giving a final answer.
• 

  liveBookmanning.com
  WordPress Power Toolkit: Harness AI to build next-level websitesChain-of-thought reasoning is a problem-solving technique in which an AI model generates intermediate reasoning steps before producing a final answer.
• 

  liveBookmanning.com
  Build AI-Enhanced Web AppsA prompting technique that guides a language model through a single sequential reasoning path.
• 

  liveBookmanning.com
  Prompt Engineering in PracticeA reasoning or prompting technique in which a model generates intermediate steps toward an answer.
• 

  liveBookmanning.com
  Grokking AI ApplicationsA prompting technique that asks a language model to express intermediate reasoning steps before presenting a conclusion.
• 

  liveBookmanning.com
  AI Agents and Applications
• 

  liveBookmanning.com
  Generative AI in ActionChain of thought is a prompting technique that decomposes a problem into intermediate reasoning steps before producing an answer.
• 

  liveBookmanning.com
  Sutskever's ListChain-of-thought prompting is a prompting technique that elicits explicit intermediate reasoning steps before a model produces its final answer.
• 

  liveBookmanning.com
  Building Reliable AI SystemsChain-of-thought prompting is a prompting method that elicits explicit intermediate reasoning steps from a model.
• 

  liveBookmanning.com
  Time Series Forecasting Using Foundation ModelsChain-of-thought prompting is a prompting technique that elicits intermediate reasoning steps before a model produces an answer.
• 

  liveBookmanning.com
  Transformers in ActionChain-of-thought prompting guides a language model by providing or eliciting a sequence of intermediate reasoning steps before the final answer.
• 

  liveBookmanning.com
  Reinforcement Learning from Human FeedbackA reasoning trace is a sequence of generated textual intermediate steps used while solving a problem before reaching an answer.
• 

  liveBookmanning.com
  Deep Learning with Python, Third EditionA search-based inference approach that explores reasoning in natural language by generating intermediate thought steps before an answer.

Compare how 18 books explain chain-of-thought prompting →

Related concepts

• prompt engineeringkind of · Prompt engineering is the deliberate design and refinement of inputs, instructions, context, or examples to guide an AI model's behavior or outputs.
• language modeluses · A language model is a computational model that represents patterns in linguistic tokens to estimate their probabilities and predict or generate sequences of them.
• multi-step reasoninguses · Multi-step reasoning derives a conclusion, solution, or decision through multiple connected intermediate inferential or problem-solving steps.
• Tree-of-thoughtcontrasts with · Tree-of-thought is a reasoning technique that explores and evaluates multiple branching paths of intermediate reasoning to develop a solution.
• self-consistencyused by · Self-consistency is a technique that generates multiple reasoning paths or responses for the same input and selects or combines results based on their agreement.

This explainer was generated by Manning's knowledge graph from the text of How Large Language Models Work. The section links point to the parts of the book it draws on.