Thinking with Reasoning Skills: Fewer Tokens, More Accuracy
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911622431244288 |
|---|---|
| author | Zhao, Guangxiang Shi, Qilong Xiao, Xusen Zhang, Xiangzheng Yang, Tong Sun, Lin |
| author_facet | Zhao, Guangxiang Shi, Qilong Xiao, Xusen Zhang, Xiangzheng Yang, Tong Sun, Lin |
| contents | Reasoning LLMs often spend substantial tokens on long intermediate reasoning traces (e.g., chain-of-thought) when solving new problems. We propose to summarize and store reusable reasoning skills distilled from extensive deliberation and trial-and-error exploration, and to retrieve these skills at inference time to guide future reasoning. Unlike the prevailing \emph{reasoning from scratch} paradigm, our approach first recalls relevant skills for each query, helping the model avoid redundant detours and focus on effective solution paths. We evaluate our method on coding and mathematical reasoning tasks, and find that it significantly reduces reasoning tokens while improving overall performance. The resulting lower per-request cost indicates strong practical and economic potential for real-world deployment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_21764 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Thinking with Reasoning Skills: Fewer Tokens, More Accuracy Zhao, Guangxiang Shi, Qilong Xiao, Xusen Zhang, Xiangzheng Yang, Tong Sun, Lin Artificial Intelligence Reasoning LLMs often spend substantial tokens on long intermediate reasoning traces (e.g., chain-of-thought) when solving new problems. We propose to summarize and store reusable reasoning skills distilled from extensive deliberation and trial-and-error exploration, and to retrieve these skills at inference time to guide future reasoning. Unlike the prevailing \emph{reasoning from scratch} paradigm, our approach first recalls relevant skills for each query, helping the model avoid redundant detours and focus on effective solution paths. We evaluate our method on coding and mathematical reasoning tasks, and find that it significantly reduces reasoning tokens while improving overall performance. The resulting lower per-request cost indicates strong practical and economic potential for real-world deployment. |
| title | Thinking with Reasoning Skills: Fewer Tokens, More Accuracy |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2604.21764 |