Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2502.18600 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909522228936704 |
|---|---|
| author | Xu, Silei Xie, Wenhao Zhao, Lingxiao He, Pengcheng |
| author_facet | Xu, Silei Xie, Wenhao Zhao, Lingxiao He, Pengcheng |
| contents | Large Language Models (LLMs) have demonstrated remarkable performance in solving complex reasoning tasks through mechanisms like Chain-of-Thought (CoT) prompting, which emphasizes verbose, step-by-step reasoning. However, humans typically employ a more efficient strategy: drafting concise intermediate thoughts that capture only essential information. In this work, we propose Chain of Draft (CoD), a novel paradigm inspired by human cognitive processes, where LLMs generate minimalistic yet informative intermediate reasoning outputs while solving tasks. By reducing verbosity and focusing on critical insights, CoD matches or surpasses CoT in accuracy while using as little as only 7.6% of the tokens, significantly reducing cost and latency across various reasoning tasks. Our code and data are available at https://github.com/sileix/chain-of-draft. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_18600 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Chain of Draft: Thinking Faster by Writing Less Xu, Silei Xie, Wenhao Zhao, Lingxiao He, Pengcheng Computation and Language I.2.7 Large Language Models (LLMs) have demonstrated remarkable performance in solving complex reasoning tasks through mechanisms like Chain-of-Thought (CoT) prompting, which emphasizes verbose, step-by-step reasoning. However, humans typically employ a more efficient strategy: drafting concise intermediate thoughts that capture only essential information. In this work, we propose Chain of Draft (CoD), a novel paradigm inspired by human cognitive processes, where LLMs generate minimalistic yet informative intermediate reasoning outputs while solving tasks. By reducing verbosity and focusing on critical insights, CoD matches or surpasses CoT in accuracy while using as little as only 7.6% of the tokens, significantly reducing cost and latency across various reasoning tasks. Our code and data are available at https://github.com/sileix/chain-of-draft. |
| title | Chain of Draft: Thinking Faster by Writing Less |
| topic | Computation and Language I.2.7 |
| url | https://arxiv.org/abs/2502.18600 |