Chain-of-thought (CoT)
Düşünce zinciri (CoT)
D3 · Prompts, context, RAG, tools and agents
A chain of thought is a sequence of intermediate reasoning steps expressed while working toward an answer.
Review status: 2026-12-05
Technical explanation
Chain-of-thought prompting supplies examples of intermediate steps to guide the model’s response.
Conceptual boundaries
A generated reasoning trace is an output representation; it is not direct access to all internal computation.
Provider-neutral example
Illustrative: a worked arithmetic answer separates identifying quantities, choosing an operation and checking the result.
Limitations
Intermediate steps can be misleading or incomplete, so they do not replace checking the answer.
Related concepts
Atomic claims and evidence
1.1Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
- Source
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
- Source role
- Authoritative source
- Exact locator
- Abstract: intermediate reasoning steps and chain-of-thought demonstrations
- Supported claim
- Wei and colleagues describe CoT as intermediate reasoning steps and test prompting with worked demonstrations.
- Last verification
- Review due
- Scope limitation
- The paper studies particular models and reasoning tasks.
2.1Anthropic: Reasoning models do not always say what they think
- Source
- Anthropic: Reasoning models do not always say what they think
- Source role
- Authoritative source
- Exact locator
- Results: models can use hints without acknowledging them in their chain of thought
- Supported claim
- Anthropic’s experiments show that reasoning models can use provided hints without acknowledging them in their reasoning text.
- Last verification
- Review due
- Scope limitation
- This is evidence of possible unfaithfulness, not proof that every trace is false.