liveBook · Manning

Chain-of-thought prompting

As explained in Machines that Think

A prompting technique that encourages a model to produce intermediate reasoning steps before its final answer.

Chain-of-thought prompting asks a model to produce intermediate reasoning steps in natural language before giving its final answer. This added structure can improve performance on tasks requiring multiple steps, such as arithmetic or logical reasoning, but does not guarantee correctness. section 6.2.1 section 9.6

Interpret these explanations cautiously:

  • Treat them as generated outputs, not guaranteed records of the factors that caused an answer. Hidden prompt biases can influence a model’s answer without appearing in its plausible-sounding explanation. section 7.2.1
  • Distinguish intermediate reasoning from verification: chain-of-verification adds a separate check after an initial answer, but can still reproduce or rationalize errors. section 6.2.1
Figure 6.4 Structuring generation through reasoning and verification. In a direct approach, the model produces an answer in a single step. Chain-of-thought (CoT) prompting introduces intermediate reasoning steps that make the path to the answer explicit. Chain-of-verification (CoV) extends this process by adding a second stage in which the initial answer is checked against supporting conditions, reducing the likelihood of unsupported conclusions.

Compare direct answering, intermediate reasoning, and a separate verification stage. section 6.2.1

How Machines that Think covers it

How deep it goes
Defines the prompting approach and its rationale, then qualifies its limits without detailing implementation.
Examples
Uses arithmetic and logical reasoning tasks, plus hidden prompt biases that produce plausible but unfaithful explanations.
What you can do after
Evaluate reasoning explanations cautiously rather than treating them as faithful accounts of a model’s causal process.
How it connects
Links chain-of-thought prompting to chain-of-verification, hallucination, alignment, and reasoning models.

Also explained in

Compare how 29 books explain chain-of-thought prompting →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machines that Think. The section links point to the parts of the book it draws on.