Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.08324 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913160917680128 |
|---|---|
| author | Tang, Yuntian Jia, Bohan Huang, Wenxuan Zhang, Lianyue Xie, Jiao Li, Wenxi Li, Wei Hu, Jie Ji, Xinghao Chen Rongrong Lin, Shaohui |
| author_facet | Tang, Yuntian Jia, Bohan Huang, Wenxuan Zhang, Lianyue Xie, Jiao Li, Wenxi Li, Wei Hu, Jie Ji, Xinghao Chen Rongrong Lin, Shaohui |
| contents | Chain-of-Thought (CoT) reasoning successfully enhances the reasoning capabilities of Large Language Models (LLMs), yet it incurs substantial computational overhead for inference. Existing CoT compression methods often suffer from a critical loss of logical fidelity at high compression ratios, resulting in significant performance degradation. To achieve high-fidelity, fast reasoning, we propose a novel EXTreme-RAtio Chain-of-Thought Compression framework, termed Extra-CoT, which aggressively reduces the token budget while preserving answer accuracy. To generate reliable, high-fidelity supervision, we first train a dedicated semantically-preserved compressor on mathematical CoT data with fine-grained annotations. An LLM is then fine-tuned on these compressed pairs via a mixed-ratio supervised fine-tuning (SFT), teaching it to follow a spectrum of compression budgets and providing a stable initialization for reinforcement learning (RL). We further propose Constrained and Hierarchical Ratio Policy Optimization (CHRPO) to explicitly incentivize question-solving ability under lower budgets by a hierarchical reward. Experiments on three mathematical reasoning benchmarks show the superiority of Extra-CoT. For example, on MATH-500 using Qwen3-1.7B, Extra-CoT achieves over 73\% token reduction with an accuracy improvement of 0.6\%, significantly outperforming state-of-the-art (SOTA) methods. Our source codes have been released at https://github.com/Mwie1024/Extra-CoT. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_08324 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression Tang, Yuntian Jia, Bohan Huang, Wenxuan Zhang, Lianyue Xie, Jiao Li, Wenxi Li, Wei Hu, Jie Ji, Xinghao Chen Rongrong Lin, Shaohui Machine Learning Chain-of-Thought (CoT) reasoning successfully enhances the reasoning capabilities of Large Language Models (LLMs), yet it incurs substantial computational overhead for inference. Existing CoT compression methods often suffer from a critical loss of logical fidelity at high compression ratios, resulting in significant performance degradation. To achieve high-fidelity, fast reasoning, we propose a novel EXTreme-RAtio Chain-of-Thought Compression framework, termed Extra-CoT, which aggressively reduces the token budget while preserving answer accuracy. To generate reliable, high-fidelity supervision, we first train a dedicated semantically-preserved compressor on mathematical CoT data with fine-grained annotations. An LLM is then fine-tuned on these compressed pairs via a mixed-ratio supervised fine-tuning (SFT), teaching it to follow a spectrum of compression budgets and providing a stable initialization for reinforcement learning (RL). We further propose Constrained and Hierarchical Ratio Policy Optimization (CHRPO) to explicitly incentivize question-solving ability under lower budgets by a hierarchical reward. Experiments on three mathematical reasoning benchmarks show the superiority of Extra-CoT. For example, on MATH-500 using Qwen3-1.7B, Extra-CoT achieves over 73\% token reduction with an accuracy improvement of 0.6\%, significantly outperforming state-of-the-art (SOTA) methods. Our source codes have been released at https://github.com/Mwie1024/Extra-CoT. |
| title | Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2602.08324 |