# Chain of Draft

> Chain of Draft is a variant of chain-of-thought prompting in which the model writes only a minimal draft of each reasoning step, a few words rather than full sentences, to reach an answer with far fewer tokens.

- Identifier: PTL-0116
- Category: Reasoning Elicitation
- Canonical URL: https://protologue.com/t/chain-of-draft/
- Also known as: CoD, concise chain-of-thought
- Introduced: 2025

## Description

The authors report accuracy comparable to standard chain-of-thought on arithmetic, commonsense, and symbolic tasks while using a small fraction of the output tokens, which lowers cost and latency. It is most useful where reasoning helps but verbose explanations are not needed.

## Example

```
Think step by step, but keep only a minimum draft for each step, five words at most. Give the answer after ####.
```

## Broader terms

- [Chain-of-Thought Prompting](https://protologue.com/t/chain-of-thought/)

## Related terms

- [Prompt Compression](https://protologue.com/t/prompt-compression/)
- [Skeleton-of-Thought](https://protologue.com/t/skeleton-of-thought/)
- [Overthinking](https://protologue.com/t/overthinking/)

## Sources

- Xu et al. (2025). Chain of Draft: Thinking Faster by Writing Less. https://arxiv.org/abs/2502.18600

## Cite this entry

Protologue. (2026). Chain of Draft. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0116). https://protologue.com/t/chain-of-draft/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)