Chain-of-thought
A pattern in which a model produces intermediate reasoning steps on the way to an answer, improving accuracy on problems that require multiple dependent steps.
In more detail
Chain-of-thought refers to a model reasoning through a problem in steps rather than jumping to a conclusion. For questions that require several dependent inferences, arithmetic, or careful constraint-checking, working through the intermediate steps tends to produce more reliable answers than a single-shot response.
The idea has moved from a prompting trick into models trained to reason this way by default. Either way the principle holds: giving a model room to work through a hard problem before committing to an answer improves reliability, at the cost of extra inference time.
Where this shows up at Ceven
Ceven applies step-by-step reasoning where it pays off, such as planning how to fulfill a plain-language request or working through a research question, while keeping simple steps fast. Deep research in particular depends on careful multi-step reasoning to move from a pile of sources to a coherent, cited brief.