Elamur Research⌕
☰
home / papers / 2025 / Dynamic Early Exit in Reasoning Models

Dynamic Early Exit in Reasoning Models

Paper · 2025 · Chenxu Yang, Qingyi Si, Yongjie Duan et al.

Summary

Large reasoning models such as DeepSeek-R1 and GPT-O1 often produce very long chains of thought (CoT). This "overthinking" adds latency and compute cost, and it can also turn a correct line of reasoning into a wrong one. In pilot runs, the authors forced models to answer at different points in their reasoning. Many problems could already be solved partway through, and the best stopping point differed from problem to problem. Fixed cut-off rules are therefore a poor fit.

They propose DEER (Dynamic Early Exit in Reasoning), which needs no training. While the model generates, DEER looks for moments where it switches thoughts. It detects these either by linguistic markers like "Wait" or by high entropy in the first token of a new step. At each such moment, it inserts a "final answer" prompt to obtain a trial answer. If the model is confident in that answer, reasoning stops. If not, the probe is discarded and generation continues as before. A variant, DEER-Pro, runs several answer probes in parallel and combines them for a more stable confidence estimate.

This gives a plug-and-play way to make reasoning models both faster and more accurate.

Key points

  • Results: tested on 10 benchmarks (e.g., GSM8K, MATH-500, AMC, GPQA Diamond, AIME, LiveCodeBench, BigCodeBench) and 11 reasoning LLMs of different families and sizes. On average, CoT length fell by 19.1%–80.1% and accuracy rose by 0.3%–5.0%.
  • "Pearl reasoning" on AIME24: about 75% of samples had an earlier stopping point that still gave a correct answer. In 36.7% of samples, less than half of the original reasoning was enough.
  • Stopping after only 20% of reasoning steps: 60.8% of originally correct MATH-500 answers stayed correct, as did 35.1% on GPQA.
  • Fixing wrong answers: early exit corrected the most errors at about 40% of the steps on MATH and about 50% on GPQA.
  • Code: available at github.com/iie-ycx/DEER.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏛️Chinese Academy of SciencesBeijing · China246 works · 24,951 citations🎓University of Chinese Academy of SciencesBeijing · China188 works · 8,743 citations🔬Institute of Information EngineeringBeijing · China41 works · 1,485 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,685 works

Cite

Cite as Chenxu Yang, Qingyi Si, Yongjie Duan, Zheliang Zhu, Chenyu Zhu, Qiaowei Li, Minghui Chen, Zheng Lin and Weiping Wang. 2025. Dynamic Early Exit in Reasoning Models. arXiv:2504.15895.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2504.15895 · arXiv 2504.15895

@misc{yang2025dynamic,
  title = {{Dynamic Early Exit in Reasoning Models}},
  author = {Chenxu Yang, Qingyi Si, Yongjie Duan, Zheliang Zhu, Chenyu Zhu, Qiaowei Li, Minghui Chen, Zheng Lin and Weiping Wang},
  year = {2025},
  doi = {10.48550/arxiv.2504.15895},
  eprint = {2504.15895},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2504.15895}
}

References

96 references, 49 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023.
  2. Pranjal Aggarwal and Sean Welleck. L1: Controlling how long a reasoning model thinks with reinforcement learning. arXiv preprint arXiv:2503.04697, 2025.
  3. Pranjal Aggarwal, Aman Madaan, Yiming Yang, et al. Let’s sample step by step: Adaptive-consistency for efficient reasoning and coding with llms. arXiv preprint arXiv:2305.11860, 2023. find
  4. AI-MO. Amc 2023, 2024. aimo-validation-amc. find
  5. URL — Training Language Models to Reason Efficiently https://huggingface.co/datasets/AI-MO/Daman Arora and Andrea Zanette. Training language models to reason efficiently. arXiv preprint arXiv:2502.04463, 2025. ↗
  6. Simon A Aytes, Jinheon Baek, and Sung Ju Hwang. Sketch-of-thought: Efficient llm reasoning with adaptive cognitive-inspired sketching. arXiv preprint arXiv:2503.05179, 2025.
  7. Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, et al. Training a helpful and harmless assistant with reinforcement learning from human feedback. arXiv preprint arXiv:2204.05862, 2022. find
  8. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhai… — Evaluating Large Language Models Trained on Code
  9. Qiguang Chen, Libo Qin, Jiaqi Wang, Jingxuan Zhou, and Wanxiang Che. Unlocking the capabilities of thought: A reasoning boundary framework to quantify and optimize chain-of-thought. Advances in Neural Information Processing Systems, 37:54872–54904, 2024. find
  10. Runjin Chen, Zhenyu Zhang, Junyuan Hong, Souvik Kundu, and Zhangyang Wang. Seal: Steerable reasoning calibration of large language models for free, 2025a. URL https://arxiv.org/ abs/2504.07986. — [arXiv:2504.07986](https://arxiv.org/abs/2504.07986) ↗
  11. Xingyu Chen, Jiahao Xu, Tian Liang, Zhiwei He, Jianhui Pang, Dian Yu, Linfeng Song, Qiuzhi Liu, Mengfei Zhou, Zhuosheng Zhang, Rui Wang, Zhaopeng Tu, Haitao Mi, and Dong Yu. Do not think that much for 2+3=? on the overthinking of o1-like llms, 2025b. URL https://arxiv.org/abs/2412.21187. ↗
  12. Jeffrey Cheng and Benjamin Van Durme. Compressed chain of thought: Efficient reasoning through dense representations. arXiv preprint arXiv:2412.13171, 2024.
  13. Yu-Neng Chuang, Helen Zhou, Prathusha Sarma, Parikshit Gopalan, John Boccio, Sara Bolouki, and Xia Hu. Learning to route llms with confidence tokens. arXiv preprint arXiv, 2410, 2024. find
  14. Yu-Neng Chuang, Leisheng Yu, Guanchu Wang, Lizhe Zhang, Zirui Liu, Xuanting Cai, Yang Sui, Vladimir Braverman, and Xia Hu. Confident or seek stronger: Exploring uncertainty-based on-device llm routing from benchmarking to generalization. arXiv preprint arXiv:2502.04428, 2025. — [arXiv:2502.04428](https://arxiv.org/abs/2502.04428) find
  15. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. Training verifiers to solve math word problems, 2021. URL https://arxiv.org/abs/2110.14168. ↗
  16. Alejandro Cuadron, Dacheng Li, Wenjie Ma, Xingyao Wang, Yichuan Wang, Siyuan Zhuang, Shu Liu, Luis Gaspar Schroeder, Tian Xia, Huanzhi Mao, et al. The danger of overthinking: Examining the reasoning-action dilemma in agentic tasks. arXiv preprint arXiv:2502.08235, 2025. find
  17. Yingqian Cui, Pengfei He, Jingying Zeng, Hui Liu, Xianfeng Tang, Zhenwei Dai, Yan Han, Chen Luo, Jing Huang, Zhen Li, et al. Stepwise perplexity-guided refinement for efficient chain-of-thought reasoning in large language models. arXiv preprint arXiv:2502.13260, 2025. find
  18. Muzhi Dai, Chenxu Yang, and Qingyi Si. S-grpo: Early exit via reinforcement learning in reasoning models, 2025. URL https://arxiv.org/abs/2505.07686. — [arXiv:2505.07686](https://arxiv.org/abs/2505.07686) ↗
  19. Renfei Dang, Shujian Huang, and Jiajun Chen. Internal bias in reasoning models leads to overthinking, 2025. URL https://arxiv.org/abs/2505.16448. — [arXiv:2505.16448](https://arxiv.org/abs/2505.16448) ↗
  20. DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei Feng, Chengda Lu, Chengg… — DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
  21. Razvan-Gabriel Dumitru, Darius Peteleaza, Vikas Yadav, and Liangming Pan. Conciserl: Conciseness-guided reinforcement learning for efficient reasoning models, 2025. URL https://arxiv.org/abs/2505.17250. ↗
  22. Yichao Fu, Junda Chen, Yonghao Zhuang, Zheyu Fu, Ion Stoica, and Hao Zhang. Reasoning without self-doubt: More efficient chain-of-thought through certainty probing. In ICLR 2025 Workshop on Foundation Models in the Wild, 2025. URL https://openreview.net/forum?id= wpK4IMJfdX. ↗
  23. Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyu Zhao, Shiqing Ma, and Zhenyu Chen. Token-budget-aware llm reasoning. arXiv preprint arXiv:2412.18547, 2024.
  24. Shibo Hao, Sainbayar Sukhbaatar, DiJia Su, Xian Li, Zhiting Hu, Jason Weston, and Yuandong Tian. Training large language models to reason in a continuous latent space. arXiv preprint arXiv:2412.06769, 2024. find
  25. Chaoqun He, Renjie Luo, Yuzhuo Bai, Shengding Hu, Zhen Leng Thai, Junhao Shen, Jinyi Hu, Xu Han, Yujie Huang, Yuxiang Zhang, Jie Liu, Lei Qi, Zhiyuan Liu, and Maosong Sun. Olympiadbench: A challenging benchmark for promoting agi with olympiad-level bilingual multimodal scientific problems, 2024. URL…
  26. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. Measuring mathematical problem solving with the math dataset, 2021. URL https://arxiv.org/abs/2103.03874. ↗
  27. Yao Huang, Huanran Chen, Shouwei Ruan, Yichi Zhang, Xingxing Wei, and Yinpeng Dong. Mitigating overthinking in large reasoning models via manifold steering, 2025. URL https://arxiv.org/abs/2505.22411. — [arXiv:2505.22411](https://arxiv.org/abs/2505.22411) ↗
  28. Naman Jain, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, and Ion Stoica. Livecodebench: Holistic and contamination free evaluation of large language models for code, 2024. URL https://arxiv.org/abs/ 2403.07974. ↗
  29. Lingjie Jiang, Xun Wu, Shaohan Huang, Qingxiu Dong, Zewen Chi, Li Dong, Xingxing Zhang, Tengchao Lv, Lei Cui, and Furu Wei. Think only when you need with large hybrid-reasoning models, 2025a. URL https://arxiv.org/abs/2505.14631. — [arXiv:2505.14631](https://arxiv.org/abs/2505.14631) ↗
  30. Yuxuan Jiang, Dawei Li, and Frank Ferraro. Drp: Distilled reasoning pruning with skill-aware step decomposition for efficient large reasoning models, 2025b. URL https://arxiv.org/ abs/2505.13975. ↗
  31. Yu Kang, Xianghui Sun, Liangyu Chen, and Wei Zou. C3ot: Generating shorter chain-of-thought without compromising effectiveness. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pp. 24312–24320, 2025. kvcache ai. KTransformers: A flexible framework for experiencing cutting…
  32. Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention, 2023. URL https://arxiv.org/abs/2309.06180. ↗
  33. Ayeong Lee, Ethan Che, and Tianyi Peng. How well do llms compress their own chain-of-thought? a token complexity approach. arXiv preprint arXiv:2503.01141, 2025. — [arXiv:2503.01141](https://arxiv.org/abs/2503.01141) find
  34. Yiwei Li, Peiwen Yuan, Shaoxiong Feng, Boyuan Pan, Xinglin Wang, Bin Sun, Heda Wang, and Kan Li. Escape sky-high cost: Early-stopping self-consistency for multi-step reasoning. arXiv preprint arXiv:2401.10480, 2024. find
  35. Zheng Li, Qingxiu Dong, Jingyuan Ma, Di Zhang, Kai Jia, and Zhifang Sui. Selfbudgeter: Adaptive token allocation for efficient llm reasoning, 2025a. URL https://arxiv.org/abs/2505. 11274. ↗
  36. Zhong-Zhi Li, Duzhen Zhang, Ming-Liang Zhang, Jiaxin Zhang, Zengyan Liu, Yuxuan Yao, Haotian Xu, Junhao Zheng, Pei-Jie Wang, Xiuyi Chen, Yingying Zhang, Fei Yin, Jiahua Dong, Zhijiang Guo, Le Song, and Cheng-Lin Liu. From system 1 to system 2: A survey of reasoning large language models, 2025b. URL…
  37. Baohao Liao, Yuhui Xu, Hanze Dong, Junnan Li, Christof Monz, Silvio Savarese, Doyen Sahoo, and Caiming Xiong. Reward-guided speculative decoding for efficient llm reasoning. arXiv preprint arXiv:2501.19324, 2025. — [arXiv:2501.19324](https://arxiv.org/abs/2501.19324) find
  38. Hanbing Liu, Lang Cao, Yuanyi Ren, Mengyu Zhou, Haoyu Dong, Xiaojun Ma, Shi Han, and Dongmei Zhang. Bingo: Boosting efficient reasoning of llms via dynamic and significance-based reinforcement learning, 2025a. URL https://arxiv.org/abs/2506.08125. — [arXiv:2506.08125](https://arxiv.org/abs/2506.08125) ↗
  39. Tengxiao Liu, Qipeng Guo, Xiangkun Hu, Cheng Jiayang, Yue Zhang, Xipeng Qiu, and Zheng Zhang. Can language models learn to skip steps? arXiv preprint arXiv:2411.01855, 2024. — [arXiv:2411.01855](https://arxiv.org/abs/2411.01855) find
  40. Wei Liu, Ruochen Zhou, Yiyun Deng, Yuzhen Huang, Junteng Liu, Yuntian Deng, Yizhe Zhang, and Junxian He. Learn to reason efficiently with adaptive length-based reward shaping, 2025b. URL https://arxiv.org/abs/2505.15612. — [arXiv:2505.15612](https://arxiv.org/abs/2505.15612) ↗
  41. Zichen Liu, Changyu Chen, Wenjun Li, Penghui Qi, Tianyu Pang, Chao Du, Wee Sun Lee, and Min Lin. Understanding r1-zero-like training: A critical perspective, 2025c. URL https://arxiv.org/abs/2503.20783. ↗
  42. Haotian Luo, Li Shen, Haiying He, Yibo Wang, Shiwei Liu, Wei Li, Naiqiang Tan, Xiaochun Cao, and Dacheng Tao. O1-pruner: Length-harmonizing fine-tuning for o1-like reasoning pruning. arXiv preprint arXiv:2501.12570, 2025a. find
  43. Yijia Luo, Yulin Song, Xingyao Zhang, Jiaheng Liu, Weixun Wang, GengRu Chen, Wenbo Su, and Bo Zheng. Deconstructing long chain-of-thought: A structured reasoning optimization framework for long cot distillation, 2025b. URL https://arxiv.org/abs/2503.16385. ↗
  44. Wenjie Ma, Jingxuan He, Charlie Snell, Tyler Griggs, Sewon Min, and Matei Zaharia. Reasoning models can be effective without thinking. arXiv preprint arXiv:2504.09858, 2025a. find
  45. Xinyin Ma, Guangnian Wan, Runpeng Yu, Gongfan Fang, and Xinchao Wang. Cot-valve: Length-compressible chain-of-thought tuning. arXiv preprint arXiv:2502.09601, 2025b.
  46. Rohin Manvi, Anikait Singh, and Stefano Ermon. Adaptive inference-time compute: Llms can predict if they can do better, even mid-generation. arXiv preprint arXiv:2410.02725, 2024.
  47. MAA Committees. Aime problems and solutions. — s1: Simple test-time scaling https://artofproblemsolving.com/ wiki/index.php/AIME_Problems_and_Solutions. Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Candès, and Tatsunori Hashimoto. s1: Simple… ↗
  48. Tergel Munkhbat, Namgyu Ho, Seo Hyun Kim, Yongjin Yang, Yujin Kim, and Se-Young Yun. Self-training elicits concise reasoning in large language models. arXiv preprint arXiv:2502.20122, 2025. find
  49. Isaac Ong, Amjad Almahairi, Vincent Wu, Wei-Lin Chiang, Tianhao Wu, Joseph E Gonzalez, M Waleed Kadous, and Ion Stoica. Routellm: Learning to route llms with preference data, 2024. URL https://arxiv.org/abs/2406.18665. ↗
  50. OpenAI. Learning to reason with llms. https://openai.com/research/ learning-to-reason-with-llms, 2025. Accessed: 15 March 2025. ↗
  51. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35: 27730–27744, 2022.
  52. Ziqing Qiao, Yongheng Deng, Jiali Zeng, Dong Wang, Lai Wei, Guanbo Wang, Fandong Meng, Jie Zhou, Ju Ren, and Yaoxue Zhang. Concise: Confidence-guided compression in step-by-step efficient reasoning, 2025. URL https://arxiv.org/abs/2505.04881. ↗
  53. Yuxiao Qu, Matthew YR Yang, Amrith Setlur, Lewis Tunstall, Edward Emanuel Beeching, Ruslan Salakhutdinov, and Aviral Kumar. Optimizing test-time compute via meta reinforcement fine-tuning. arXiv preprint arXiv:2503.07572, 2025. find
  54. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, Kai Dang, Keming Lu, Keqin Bao, Kexin Yang, Le Yu, Mei Li, Mingfeng… find
  55. Shyam Sundhar Ramesh, Yifan Hu, Iason Chaimalas, Viraj Mehta, Pier Giuseppe Sessa, Haitham Bou Ammar, and Ilija Bogunovic. Group robust preference optimization in reward-free RLHF. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024. URL https://openreview.net/forum… ↗
  56. Carl Rasmussen and Zoubin Ghahramani. Occam's razor. In T. Leen, T. Dietterich, and V. Tresp (eds.), Advances in Neural Information Processing Systems, volume 13. MIT Press, 2000. URL https://proceedings.neurips.cc/paper_files/paper/2000/ file/0950ca92a4dcf426067cfd2246bb5ff3-Paper.pdf. ↗
  57. David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R. Bowman. Gpqa: A graduate-level google-proof q&a benchmark, 2023. URL https://arxiv.org/abs/2311.12022. ↗
  58. Matthew Renze and Erhan Guven. The benefits of a concise chain of thought on problem-solving in large language models. In 2024 2nd International Conference on Foundation and Large Language Models (FLLM), pp. 476–483. find
  59. Nikunj Saunshi, Nishanth Dikkala, Zhiyuan Li, Sanjiv Kumar, and Sashank J Reddi. Reasoning with latent thoughts: On the power of looped transformers. arXiv preprint arXiv:2502.17416, 2025. find
  60. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017. find
  61. Xuan Shen, Yizhou Wang, Xiangxi Shi, Yanzhi Wang, Pu Zhao, and Jiuxiang Gu. Efficient reasoning with hidden thinking. arXiv preprint arXiv:2501.19201, 2025a. — [arXiv:2501.19201](https://arxiv.org/abs/2501.19201) find
  62. Yi Shen, Jian Zhang, Jieyun Huang, Shuming Shi, Wenjing Zhang, Jiangze Yan, Ning Wang, Kai Wang, and Shiguo Lian. Dast: Difficulty-adaptive slow-thinking for large reasoning models. arXiv preprint arXiv:2503.04472, 2025b. find
  63. Zhenyi Shen, Hanqi Yan, Linhai Zhang, Zhanghao Hu, Yali Du, and Yulan He. Codi: Compressing chain-of-thought into continuous space via self-distillation. arXiv preprint arXiv:2502.21074, 2025c.
  64. Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. Scaling llm test-time compute optimally can be more effective than scaling model parameters, 2024. URL https://arxiv.org/ abs/2408.03314. ↗
  65. Jiwon Song, Dongwon Jo, Yulhwa Kim, and Jae-Joon Kim. Reasoning path compression: Compressing generation trajectories for efficient llm reasoning, 2025. URL https://arxiv.org/ abs/2505.13866. — [arXiv:2505.13866](https://arxiv.org/abs/2505.13866) ↗
  66. DiJia Su, Hanlin Zhu, Yingchen Xu, Jiantao Jiao, Yuandong Tian, and Qinqing Zheng. Token assorted: Mixing latent and text tokens for improved language model reasoning. arXiv preprint arXiv:2502.03275, 2025. — [arXiv:2502.03275](https://arxiv.org/abs/2502.03275) find
  67. Yang Sui, Yu-Neng Chuang, Guanchu Wang, Jiamu Zhang, Tianyi Zhang, Jiayi Yuan, Hongyi Liu, Andrew Wen, Shaochen Zhong, Hanjie Chen, and Xia Hu. Stop overthinking: A survey on efficient reasoning for large language models, 2025. URL https://arxiv.org/abs/2503. 16419. ↗
  68. Wenhui Tan, Jiaze Li, Jianzhong Ju, Zhenbo Luo, Jian Luan, and Ruihua Song. Think silently, think fast: Dynamic latent compression of llm reasoning chains, 2025. URL https://arxiv.org/ abs/2505.16552. ↗
  69. Kimi Team, Angang Du, Bofei Gao, Bowei Xing, Changjiu Jiang, Cheng Chen, Cheng Li, Chenjun Xiao, Chenzhuang Du, Chonghua Liao, Chuning Tang, Congcong Wang, Dehao Zhang, Enming Yuan, Enzhe Lu, Fengxiang Tang, Flood Sung, Guangda Wei, Guokun Lai, Haiqing Guo, Han Zhu, Hao Ding, Hao Hu, Hao Yang, Hao Z… — Kimi k1.5: Scaling Reinforcement Learning with LLMs
  70. Kimi Team, Angang Du, Bofei Gao, Bowei Xing, Changjiu Jiang, Cheng Chen, Cheng Li, Chenjun Xiao, Chenzhuang Du, Chonghua Liao, et al. Kimi k1. 5: Scaling reinforcement learning with llms. arXiv preprint arXiv:2501.12599, 2025b.
  71. Qwen Team. Qwq-32b: Embracing the power of reinforcement learning, March 2025. URL https://qwenlm.github.io/blog/qwq-32b/. ↗
  72. Songjun Tu, Jiahao Lin, Qichao Zhang, Xiangyu Tian, Linjing Li, Xiangyuan Lan, and Dongbin Zhao. Learning when to think: Shaping adaptive reasoning in r1-style models via multi-stage rl, 2025. URL https://arxiv.org/abs/2505.10832. — [arXiv:2505.10832](https://arxiv.org/abs/2505.10832) ↗
  73. Hongru WANG, Deng Cai, Wanjun Zhong, Shijue Huang, Jeff Z. Pan, Zeming Liu, and Kam-Fai Wong. Self-reasoning language models: Unfold hidden reasoning chains with few reasoning catalyst. In Workshop on Reasoning and Planning for Large Language Models, 2025. URL https://openreview.net/forum?id=p4wXiD8… ↗
  74. Rui Wang, Hongru Wang, Boyang Xue, Jianhui Pang, Shudong Liu, Yi Chen, Jiahao Qiu, Derek Fai Wong, Heng Ji, and Kam-Fai Wong. Harnessing the reasoning economy: A survey of efficient reasoning for large language models, 2025a. URL https://arxiv.org/abs/2503.24377. ↗
  75. Shenzhi Wang, Le Yu, Chang Gao, Chujie Zheng, Shixuan Liu, Rui Lu, Kai Dang, Xionghui Chen, Jianxin Yang, Zhenru Zhang, Yuqiong Liu, An Yang, Andrew Zhao, Yang Yue, Shiji Song, Bowen Yu, Gao Huang, and Junyang Lin. Beyond the 80/20 rule: High-entropy minority tokens drive effective reinforcement lea… — Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning
  76. Yibo Wang, Li Shen, Huanjin Yao, Tiansheng Huang, Rui Liu, Naiqiang Tan, Jiaxing Huang, Kai Zhang, and Dacheng Tao. R1-compress: Long chain-of-thought compression via chunk compression and search, 2025c. URL https://arxiv.org/abs/2505.16838. — [arXiv:2505.16838](https://arxiv.org/abs/2505.16838) ↗
  77. Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. Finetuned language models are zero-shot learners. arXiv preprint arXiv:2109.01652, 2021.
  78. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models, 2023. URL https://arxiv.org/abs/2201.11903. ↗
  79. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mariama Drame, Quentin Lho… — Transformers: State-of-the-Art Natural Language Processing
  80. Yuyang Wu, Yifei Wang, Tianqi Du, Stefanie Jegelka, and Yisen Wang. When more is less: Understanding chain-of-thought length in llms. arXiv preprint arXiv:2502.07266, 2025.
  81. Heming Xia, Yongqi Li, Chak Tou Leong, Wenjie Wang, and Wenjie Li. Tokenskip: Controllable chain-of-thought compression in llms. arXiv preprint arXiv:2502.12067, 2025.
  82. Yuxi Xie, Kenji Kawaguchi, Yiran Zhao, James Xu Zhao, Min-Yen Kan, Junxian He, and Michael Xie. Self-evaluation guided beam search for reasoning. Advances in Neural Information Processing Systems, 36:41618–41650, 2023.
  83. Fengli Xu, Qianyue Hao, Zefang Zong, Jingwei Wang, Yunke Zhang, Jingyi Wang, Xiaochong Lan, Jiahui Gong, Tianjian Ouyang, Fanjin Meng, Chenyang Shao, Yuwei Yan, Qinglong Yang, Yiwen Song, Sijian Ren, Xinyuan Hu, Yu Li, Jie Feng, Chen Gao, and Yong Li. Towards large reasoning models: A survey of rein… — Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models
  84. Silei Xu, Wenhao Xie, Lingxiao Zhao, and Pengcheng He. Chain of draft: Thinking faster by writing less. arXiv preprint arXiv:2502.18600, 2025b.
  85. Silei Xu, Wenhao Xie, Lingxiao Zhao, and Pengcheng He. Chain of draft: Thinking faster by writing less, 2025c. URL https://arxiv.org/abs/2502.18600. ↗
  86. Yige Xu, Xu Guo, Zhiwei Zeng, and Chunyan Miao. Softcot: Soft chain-of-thought for efficient reasoning with llms. arXiv preprint arXiv:2502.12134, 2025d.
  87. Yixin Ye, Zhen Huang, Yang Xiao, Ethan Chern, Shijie Xia, and Pengfei Liu. Limo: Less is more for reasoning, 2025. URL https://arxiv.org/abs/2502.03387. ↗
  88. Edward Yeo, Yuxuan Tong, Morry Niu, Graham Neubig, and Xiang Yue. Demystifying long chain-of-thought reasoning in llms. arXiv preprint arXiv:2502.03373, 2025.
  89. Bin Yu, Hang Yuan, Haotian Li, Xueyin Xu, Yuliang Wei, Bailing Wang, Weizhen Qi, and Kai Chen. Long-short chain-of-thought mixture supervised fine-tuning eliciting efficient reasoning in large language models, 2025. URL https://arxiv.org/abs/2505.03469. ↗
  90. Ping Yu, Jing Xu, Jason Weston, and Ilia Kulikov. Distilling system 2 into system 1. arXiv preprint arXiv:2407.06023, 2024. find
  91. Anqi Zhang, Yulin Chen, Jane Pan, Chen Zhao, Aurojit Panda, Jinyang Li, and He He. Reasoning models know when they’re right: Probing hidden states for self-verification. arXiv preprint arXiv:2504.05419, 2025a. find
  92. Jiajie Zhang, Nianyi Lin, Lei Hou, Ling Feng, and Juanzi Li. Adaptthink: Reasoning models can learn when to think, 2025b. URL https://arxiv.org/abs/2505.13417. ↗
  93. Wenyuan Zhang, Shuaiyi Nie, Xinghua Zhang, Zefeng Zhang, and Tingwen Liu. S1-bench: A simple benchmark for evaluating system 1 thinking capability of large reasoning models. ArXiv, abs/2504.10368, 2025c. URL https://api.semanticscholar.org/ CorpusID:277781494. — [arXiv:2504.10368](https://arxiv.org/abs/2504.10368) ↗
  94. Zhen Zhang, Xuehai He, Weixiang Yan, Ao Shen, Chenyang Zhao, Shuohang Wang, Yelong Shen, and Xin Eric Wang. Soft thinking: Unlocking the reasoning potential of llms in continuous concept space, 2025d. URL https://arxiv.org/abs/2505.15778. ↗
  95. Rongzhi Zhu, Yi Liu, Zequn Sun, Yiwei Wang, and Wei Hu. When can large reasoning models save thinking? mechanistic analysis of behavioral divergence in reasoning, 2025. URL https://arxiv.org/abs/2505.15276. — [arXiv:2505.15276](https://arxiv.org/abs/2505.15276) ↗
  96. Terry Yue Zhuo, Minh Chien Vu, Jenny Chim, Han Hu, Wenhao Yu, Ratnadira Widyasari, Imam Nur Bani Yusuf, Haolan Zhan, Junda He, Indraneil Paul, et al. Bigcodebench: Benchmarking code generation with diverse function calls and complex instructions. arXiv preprint arXiv:2406.15877, 2024.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yang-2025-dynamic-early-exit-reasoning-models/