{"id":27652841,"url":"https://github.com/fscdc/Awesome-Efficient-Reasoning-Models","last_synced_at":"2025-04-24T05:03:30.696Z","repository":{"id":288026485,"uuid":"964653555","full_name":"fscdc/Awesome-Efficient-Reasoning-Models","owner":"fscdc","description":"[Arxiv 2025] Efficient Reasoning Models: A Survey","archived":false,"fork":false,"pushed_at":"2025-04-22T06:20:23.000Z","size":25206,"stargazers_count":109,"open_issues_count":0,"forks_count":6,"subscribers_count":3,"default_branch":"master","last_synced_at":"2025-04-22T07:43:00.634Z","etag":null,"topics":["chain-of-thought","compression","efficient-reasoning"],"latest_commit_sha":null,"homepage":"https://arxiv.org/abs/2504.10903","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/fscdc.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-04-11T15:04:28.000Z","updated_at":"2025-04-22T06:30:46.000Z","dependencies_parsed_at":"2025-04-15T07:46:52.687Z","dependency_job_id":null,"html_url":"https://github.com/fscdc/Awesome-Efficient-Reasoning-Models","commit_stats":null,"previous_names":["fscdc/awesome-efficient-reasoning-models"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fscdc%2FAwesome-Efficient-Reasoning-Models","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fscdc%2FAwesome-Efficient-Reasoning-Models/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fscdc%2FAwesome-Efficient-Reasoning-Models/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fscdc%2FAwesome-Efficient-Reasoning-Models/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/fscdc","download_url":"https://codeload.github.com/fscdc/Awesome-Efficient-Reasoning-Models/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":250566457,"owners_count":21451231,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["chain-of-thought","compression","efficient-reasoning"],"created_at":"2025-04-24T05:01:34.115Z","updated_at":"2025-04-24T05:03:30.673Z","avatar_url":"https://github.com/fscdc.png","language":"Python","funding_links":[],"categories":["Related Surveys and Repos","Related Survey","🙏 Acknowledgments","Full List","Other Lists","Resources"],"sub_categories":["Efficient Reasoning","Published in Recent Conference/Journal","Please check out all the papers by selecting the sub-area you're interested in. On this main page, only papers released in the past 90 days are shown.","TeX Lists","Applications"],"readme":"\u003cdiv align=\"center\"\u003e\n\n  \u003ch2\u003e\u003cb\u003e Efficient Reasoning Models: A Survey \u003c/b\u003e\u003c/h2\u003e\n  \u003ch4\u003e An overview of research in efficient reasoning models\u003c/h4\u003e\n\n\u003c/div\u003e\n\n\n\u003cdiv align=\"center\"\u003e\n\n![](https://img.shields.io/github/stars/fscdc/Awesome-Efficient-Reasoning-Models?color=yellow)\n![](https://img.shields.io/github/forks/fscdc/Awesome-Efficient-Reasoning-Models?color=lightblue)\n![](https://img.shields.io/github/last-commit/fscdc/Awesome-Efficient-Reasoning-Models?color=green)\n![](https://img.shields.io/badge/PRs-Welcome-blue)\n\u003ca href=\"https://arxiv.org/abs/2504.10903\" target=\"_blank\"\u003e\u003cimg src=\"https://img.shields.io/badge/arXiv-2504.10903-009688.svg\" alt=\"arXiv\"\u003e\u003c/a\u003e\n\n\u003c/div\u003e\n\n\u003cdiv align=\"center\"\u003e\n\n**[\u003ca href=\"https://arxiv.org/abs/2504.10903\"\u003earXiv\u003c/a\u003e]** **[\u003ca href=\"https://huggingface.co/papers/2504.10903\"\u003eHuggingFace\u003c/a\u003e]**\n\n\u003c/div\u003e\n\n\n\nThis repository is for our paper:\n\n\u003e **[Efficient Reasoning Models: A Survey](https://arxiv.org/abs/2504.10903)** \\\n\u003e [Sicheng Feng](https://fscdc.github.io/)\u003csup\u003e1,2\u003c/sup\u003e, [Gongfan Fang](https://fangggf.github.io/)\u003csup\u003e1\u003c/sup\u003e, [Xinyin Ma](https://horseee.github.io/)\u003csup\u003e1\u003c/sup\u003e, [Xinchao Wang](https://sites.google.com/site/sitexinchaowang/)\u003csup\u003e1,*\u003c/sup\u003e \\\n\u003e \u003csup\u003e1\u003c/sup\u003eNational University of Singapore, Singapore \\\n\u003e \u003csup\u003e2\u003c/sup\u003eNankai University, Tianjin, China \\\n\u003e \u003csup\u003e∗\u003c/sup\u003eCorresponding author: xinchao@nus.edu.sg\n\n---\n\u003e\n\u003e 🙋 Please let us know if you find out a mistake or have any suggestions!\n\u003e \n\u003e 🌟 If you find this resource helpful, please consider to star this repository and cite our [research](#citation)!\n\n\u003cp align=\"center\"\u003e\n\u003cimg src=\"assets/figure2.svg\" width = \"95%\" alt=\"\" align=center /\u003e\n\u003c/p\u003e\n\n\n## Updates\n\n- 2025-04-16: 📝 The survey is now available on [arXiv](https://arxiv.org/abs/2504.10903)!\n- 2025-04-11: 📚 The full paper list is now available and our survey is coming soon!\n- 2025-03-16: 🚀 Efficient Reasoning Repo launched!\n\n\n## Full list\n\n\n\u003e **Contributions**\n\u003e\n\u003e If you want to add your paper or update details like conference info or code URLs, please submit a pull request. You can generate the necessary markdown for each paper by filling out `generate_item.py` and running `python generate_item.py`. We greatly appreciate your contributions. Alternatively, you can email me ([Gmail](fscnkucs@gmail.com)) the links to your paper and code, and I will add your paper to the list as soon as possible.\n\n---\n\u003cp align=\"center\"\u003e\n\u003cimg src=\"assets/taxonomy.png\" width = \"95%\" alt=\"\" align=center /\u003e\n\u003c/p\u003e\n\n### Quick Links\n  - [Make Long CoT Short](#Make-Long-CoT-Short)\n    - [SFT-based Methods](#SFT-based-Methods)\n    - [RL-based Methods](#RL-based-Methods)\n    - [Prompt-driven Methods](#Prompt-driven-Methods)\n    - [Latent Reasoning](#Latent-Reasoning)\n  - [Build SLM with Strong Reasoning Ability](#Build-SLM-with-Strong-Reasoning-Ability)\n    - [Distillation](#Distillation)\n    - [Quantization and Pruning](#Quantization-and-Pruning)\n    - [RL-based Methods](#RL-based-Methods)\n  - [Let Decoding More Efficient](#Let-Decoding-More-Efficient)\n    - [Efficient TTS](#Efficient-TTS)\n    - [Other Optimal Methods](#Other-Optimal-Methods)\n  - [Evaluation and Benchmarks](#Evaluation-and-Benchmarks)\n  - [Background Papers](#Background-Papers)\n\n\n\n\n\n### Make Long CoT Short\n\n#### SFT-based Methods\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[![Star](https://img.shields.io/github/stars/horseee/CoT-Valve.svg?style=social\u0026label=Star)](https://github.com/horseee/CoT-Valve)\u003cbr\u003e[CoT-Valve: Length-Compressible Chain-of-Thought Tuning](https://arxiv.org/abs/2502.09601) \u003cbr\u003e Xinyin Ma, Guangnian Wan, Runpeng Yu, Gongfan Fang, Xinchao Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/cot_valve.png\"\u003e |[Github](https://github.com/horseee/CoT-Valve) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.09601)|[//]: #03/16\n|[![Publish](https://img.shields.io/badge/Conference-AAAI_2025-blue)]()\u003cbr\u003e[C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness](https://arxiv.org/abs/2412.11664) \u003cbr\u003e Yu Kang, Xianghui Sun, Liangyu Chen, Wei Zou |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/co3t.png\"\u003e |[Paper](https://arxiv.org/abs/2412.11664)|[//]: #03/16\n|[![Star](https://img.shields.io/github/stars/tengxiaoliu/LM_skip.svg?style=social\u0026label=Star)](https://github.com/tengxiaoliu/LM_skip) [![Publish](https://img.shields.io/badge/Conference-NeurIPS_2024-blue)]()\u003cbr\u003e[Can Language Models Learn to Skip Steps?](https://arxiv.org/abs/2411.01855) \u003cbr\u003e Tengxiao Liu, Qipeng Guo, Xiangkun Hu, Cheng Jiayang, Yue Zhang, Xipeng Qiu, Zheng Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/skip_step.png\"\u003e |[Github](https://github.com/tengxiaoliu/LM_skip) \u003cbr\u003e [Paper](https://arxiv.org/abs/2411.01855)|[//]: #03/16\n|[Distilling System 2 into System 1](https://arxiv.org/abs/2407.06023) \u003cbr\u003e Ping Yu, Jing Xu, Jason Weston, Ilia Kulikov |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/distill_sys1_sys2.png\"\u003e |[Paper](https://arxiv.org/abs/2407.06023)|[//]: #03/16\n|[![Star](https://img.shields.io/github/stars/hemingkx/TokenSkip.svg?style=social\u0026label=Star)](https://github.com/hemingkx/TokenSkip)\u003cbr\u003e[TokenSkip: Controllable Chain-of-Thought Compression in LLMs](https://arxiv.org/abs/2502.12067) \u003cbr\u003e Heming Xia, Yongqi Li, Chak Tou Leong, Wenjie Wang, Wenjie Li |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/TokenSkip.png\"\u003e |[Github](https://github.com/hemingkx/TokenSkip) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.12067)|[//]: #03/20\n|[Stepwise Perplexity-Guided Refinement for Efficient Chain-of-Thought Reasoning in Large Language Models](https://arxiv.org/abs/2502.13260) \u003cbr\u003e Yingqian Cui, Pengfei He, Jingying Zeng, Hui Liu, Xianfeng Tang, Zhenwei Dai, Yan Han, Chen Luo, Jing Huang, Zhen Li, Suhang Wang, Yue Xing, Jiliang Tang, Qi He |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.13260v1/extracted/6214965/pics/merge.png\"\u003e |[Paper](https://arxiv.org/abs/2502.13260)| [//]: #04/08\n|[Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning](https://arxiv.org/abs/2502.18080) \u003cbr\u003e Wenkai Yang, Shuming Ma, Yankai Lin, Furu Wei |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.18080v1/x10.png\"\u003e |[Paper](https://arxiv.org/abs/2502.18080)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/TergelMunkhbat/concise-reasoning.svg?style=social\u0026label=Star)](https://github.com/TergelMunkhbat/concise-reasoning)\u003cbr\u003e[Self-Training Elicits Concise Reasoning in Large Language Models](https://arxiv.org/abs/2502.20122) \u003cbr\u003e Tergel Munkhbat, Namgyu Ho, Seo Hyun Kim, Yongjin Yang, Yujin Kim, Se-Young Yun |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.20122v2/x1.png\"\u003e |[Github](https://github.com/TergelMunkhbat/concise-reasoning) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.20122)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/GeniusHTX/TALE.svg?style=social\u0026label=Star)](https://github.com/GeniusHTX/TALE)\u003cbr\u003e[Token-Budget-Aware LLM Reasoning](https://arxiv.org/abs/2412.18547) \u003cbr\u003e Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyu Zhao, Shiqing Ma, Zhenyu Chen |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2412.18547v4/x10.png\"\u003e |[Github](https://github.com/GeniusHTX/TALE) \u003cbr\u003e [Paper](https://arxiv.org/abs/2412.18547)| [//]: #04/08\n\n\n#### RL-based Methods\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[![Star](https://img.shields.io/github/stars/StarDewXXX/O1-Pruner.svg?style=social\u0026label=Star)](https://github.com/StarDewXXX/O1-Pruner)\u003cbr\u003e[O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning](https://arxiv.org/abs/2501.12570) \u003cbr\u003e Haotian Luo, Li Shen, Haiying He, Yibo Wang, Shiwei Liu, Wei Li, Naiqiang Tan, Xiaochun Cao, Dacheng Tao |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/o1_pruner.png\"\u003e |[Github](https://github.com/StarDewXXX/O1-Pruner) \u003cbr\u003e [Paper](https://arxiv.org/abs/2501.12570)|[//]: #03/16\n|[Kimi k1.5: Scaling Reinforcement Learning with LLMs](https://arxiv.org/abs/2501.12599) \u003cbr\u003e Kimi Team |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2501.12599v2/x3.png\"\u003e |[Paper](https://arxiv.org/abs/2501.12599)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/eddycmu/demystify-long-cot.svg?style=social\u0026label=Star)](https://github.com/eddycmu/demystify-long-cot)\u003cbr\u003e[Demystifying Long Chain-of-Thought Reasoning in LLMs](https://arxiv.org/abs/2502.03373) \u003cbr\u003e Edward Yeo, Yuxuan Tong, Morry Niu, Graham Neubig, Xiang Yue |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.03373v1/x1.png\"\u003e |[Github](https://github.com/eddycmu/demystify-long-cot) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.03373)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Zanette-Labs/efficient-reasoning.svg?style=social\u0026label=Star)](https://github.com/Zanette-Labs/efficient-reasoning)\u003cbr\u003e[Training Language Models to Reason Efficiently](https://arxiv.org/abs/2502.04463) \u003cbr\u003e Daman Arora, Andrea Zanette |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.04463v2/x3.png\"\u003e |[Github](https://github.com/Zanette-Labs/efficient-reasoning) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.04463)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/cmu-l3/l1.svg?style=social\u0026label=Star)](https://github.com/cmu-l3/l1)\u003cbr\u003e[L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning](https://www.arxiv.org/abs/2503.04697) \u003cbr\u003e Pranjal Aggarwal, Sean Welleck |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.04697v1/x2.png\"\u003e |[Github](https://github.com/cmu-l3/l1) \u003cbr\u003e [Paper](https://www.arxiv.org/abs/2503.04697)| [//]: #04/08\n|[DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models](https://arxiv.org/abs/2503.04472) \u003cbr\u003e Yi Shen, Jian Zhang, Jieyun Huang, Shuming Shi, Wenjing Zhang, Jiangze Yan, Ning Wang, Kai Wang, Shiguo Lian |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.04472v1/extracted/6254851/DAST.png\"\u003e |[Paper](https://arxiv.org/abs/2503.04472)| [//]: #04/08\n|[Adaptive Group Policy Optimization: Towards Stable Training and Token-Efficient Reasoning](https://arxiv.org/abs/2503.15952) \u003cbr\u003e Chen Li, Nazhou Liu, Kai Yang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/agpo.png\"\u003e |[Paper](https://arxiv.org/abs/2503.15952)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/UCSB-NLP-Chang/ThinkPrune.svg?style=social\u0026label=Star)](https://github.com/UCSB-NLP-Chang/ThinkPrune)\u003cbr\u003e[ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning](https://arxiv.org/abs/2504.01296) \u003cbr\u003e Bairu Hou, Yang Zhang, Jiabao Ji, Yujian Liu, Kaizhi Qian, Jacob Andreas, Shiyu Chang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.01296v1/x1.png\"\u003e |[Github](https://github.com/UCSB-NLP-Chang/ThinkPrune) \u003cbr\u003e [Paper](https://arxiv.org/abs/2504.01296)| [//]: #04/08\n|[Think When You Need: Self-Adaptive Chain-of-Thought Learning](https://arxiv.org/abs/2504.03234) \u003cbr\u003e Junjie Yang, Ke Lin, Xing Yu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.03234v1/extracted/6335120/alg_illu.png\"\u003e |[Paper](https://arxiv.org/abs/2504.03234)| [//]: #04/08\n\n\n\n#### Prompt-driven Methods\n\n##### Prompt-guided Efficint Reasoning\n\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[Time's Up! An Empirical Study of LLM Reasoning Ability Under Output Length Constraint](https://arxiv.org/abs/2504.14350) \u003cbr\u003e Yi Sun, Han Wang, Jiaqiang Li, Jiacheng Liu, Xiangyu Li, Hao Wen, Huiwen Zheng, Yan Liang, Yuanchun Li, Yunxin Liu |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/time_up.png\"\u003e |[Paper](https://arxiv.org/abs/2504.14350)| [//]: #04/23\n|[CoT-RAG: Integrating Chain of Thought and Retrieval-Augmented Generation to Enhance Reasoning in Large Language Models](https://arxiv.org/abs/2504.13534) \u003cbr\u003e Feiyang Li, Peng Fang, Zhan Shi, Arijit Khan, Fang Wang, Dan Feng, Weihao Wang, Xin Zhang, Yongjian Cui |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.13534v1/x2.png\"\u003e |[Paper](https://arxiv.org/abs/2504.13534)| [//]: #04/21\n|[Thought Manipulation: External Thought Can Be Efficient for Large Reasoning Models](https://arxiv.org/abs/2504.13626) \u003cbr\u003e Yule Liu, Jingyi Zheng, Zhen Sun, Zifan Peng, Wenhan Dong, Zeyang Sha, Shiwen Cui, Weiqiang Wang, Xinlei He |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/thoughtmani.png\"\u003e |[Paper](https://arxiv.org/abs/2504.13626)| [//]: #04/21\n|[![Star](https://img.shields.io/github/stars/GeniusHTX/TALE.svg?style=social\u0026label=Star)](https://github.com/GeniusHTX/TALE)\u003cbr\u003e[Token-Budget-Aware LLM Reasoning](https://arxiv.org/abs/2412.18547) \u003cbr\u003e Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyu Zhao, Shiqing Ma, Zhenyu Chen |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2412.18547v4/x10.png\"\u003e |[Github](https://github.com/GeniusHTX/TALE) \u003cbr\u003e [Paper](https://arxiv.org/abs/2412.18547)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/matthewrenze/jhu-concise-cot.svg?style=social\u0026label=Star)](https://github.com/matthewrenze/jhu-concise-cot) [![Publish](https://img.shields.io/badge/Conference-FLLM_2024-blue)]()\u003cbr\u003e[The Benefits of a Concise Chain of Thought on Problem-Solving in Large Language Models](https://arxiv.org/abs/2401.05618) \u003cbr\u003e Matthew Renze, Erhan Guven |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2401.05618v3/x1.png\"\u003e |[Github](https://github.com/matthewrenze/jhu-concise-cot) \u003cbr\u003e [Paper](https://arxiv.org/abs/2401.05618)| [//]: #04/08\n|[Break the Chain: Large Language Models Can be Shortcut Reasoners](https://arxiv.org/abs/2406.06580) \u003cbr\u003e Mengru Ding, Hanmeng Liu, Zhizhang Fu, Jian Song, Wenbo Xie, Yue Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2406.06580v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2406.06580)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/sileix/chain-of-draft.svg?style=social\u0026label=Star)](https://github.com/sileix/chain-of-draft)\u003cbr\u003e[Chain of Draft: Thinking Faster by Writing Less](https://arxiv.org/abs/2502.18600) \u003cbr\u003e Silei Xu, Wenhao Xie, Lingxiao Zhao, Pengcheng He |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.18600v2/extracted/6244873/plot.png\"\u003e |[Github](https://github.com/sileix/chain-of-draft) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.18600)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/LightChen233/reasoning-boundary.svg?style=social\u0026label=Star)](https://github.com/LightChen233/reasoning-boundary) [![Publish](https://img.shields.io/badge/Conference-NeurIPS_2024-blue)]()\u003cbr\u003e[Unlocking the Capabilities of Thought: A Reasoning Boundary Framework to Quantify and Optimize Chain-of-Thought](https://arxiv.org/abs/2410.05695) \u003cbr\u003e Qiguang Chen, Libo Qin, Jiaqi Wang, Jinxuan Zhou, Wanxiang Che |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2410.05695v2/x1.png\"\u003e |[Github](https://github.com/LightChen233/reasoning-boundary) \u003cbr\u003e [Paper](https://arxiv.org/abs/2410.05695)| [//]: #04/08\n|[How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach](https://arxiv.org/abs/2503.01141) \u003cbr\u003e Ayeong Lee, Ethan Che, Tianyi Peng |\u003cimg src=\"https://arxiv.org/html/2503.01141v2/extracted/6325669/plot/mmlu-pro-legend.png\" width=\"45%\"\u003e \u003cimg src=\"https://arxiv.org/html/2503.01141v2/extracted/6325669/plot/Anthropic/claude-3-5-sonnet-20241022-mmlu-main.png\" width=\"45%\"\u003e |[Paper](https://arxiv.org/abs/2503.01141)| [//]: #04/08\n\n\n\n##### Prompt Attribute-Aware Reasoning Routing\n\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach](https://arxiv.org/abs/2503.01141) \u003cbr\u003e Ayeong Lee, Ethan Che, Tianyi Peng |\u003cimg src=\"https://arxiv.org/html/2503.01141v2/extracted/6325669/plot/mmlu-pro-legend.png\" width=\"45%\"\u003e \u003cimg src=\"https://arxiv.org/html/2503.01141v2/extracted/6325669/plot/Anthropic/claude-3-5-sonnet-20241022-mmlu-main.png\" width=\"45%\"\u003e |[Paper](https://arxiv.org/abs/2503.01141)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-ICLR_2025-blue)]()\u003cbr\u003e[RouteLLM: Learning to Route LLMs with Preference Data](https://arxiv.org/abs/2406.18665) \u003cbr\u003e Isaac Ong, Amjad Almahairi, Vincent Wu, Wei-Lin Chiang, Tianhao Wu, Joseph E. Gonzalez, M Waleed Kadous, Ion Stoica |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2406.18665v4/extracted/6226172/Figs/gsm8k.png\"\u003e |[Paper](https://arxiv.org/abs/2406.18665)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/SimonAytes/SoT.svg?style=social\u0026label=Star)](https://github.com/SimonAytes/SoT)\u003cbr\u003e[Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching](https://arxiv.org/abs/2503.05179) \u003cbr\u003e Simon A. Aytes, Jinheon Baek, Sung Ju Hwang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.05179v1/x1.png\"\u003e |[Github](https://github.com/SimonAytes/SoT) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.05179)| [//]: #04/08\n|[Learning to Route LLMs with Confidence Tokens](https://arxiv.org/abs/2410.13284) \u003cbr\u003e Yu-Neng Chuang, Helen Zhou, Prathusha Kameswara Sarma, Parikshit Gopalan, John Boccio, Sara Bolouki, Xia Hu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2410.13284v2/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2410.13284)| [//]: #04/08\n|[Confident or Seek Stronger: Exploring Uncertainty-Based On-device LLM Routing From Benchmarking to Generalization](https://arxiv.org/abs/2502.04428) \u003cbr\u003e Yu-Neng Chuang, Leisheng Yu, Guanchu Wang, Lizhe Zhang, Zirui Liu, Xuanting Cai, Yang Sui, Vladimir Braverman, Xia Hu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.04428v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2502.04428)| [//]: #04/08\n\n###### Blog\n* [Claude 3.7 Sonnet](https://www.anthropic.com/news/claude-3-7-sonnet). Claude team. [[Paper]](https://www.anthropic.com/news/claude-3-7-sonnet)\n\n\n#### Latent Reasoning\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[Beyond Chains of Thought: Benchmarking Latent-Space Reasoning Abilities in Large Language Models](https://arxiv.org/abs/2504.10615) \u003cbr\u003e Thilo Hagendorff, Sarah Fabi |\u003cimg width=\"1002\" alt=\"image\" src=\"./figures/BCoT.png\"\u003e |[Paper](https://arxiv.org/abs/2504.10615)|[//]: #04/17\n|[Distilling System 2 into System 1](https://arxiv.org/abs/2407.06023) \u003cbr\u003e Ping Yu, Jing Xu, Jason Weston, Ilia Kulikov |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/distill_sys1_sys2.png\"\u003e |[Paper](https://arxiv.org/abs/2407.06023)|[//]: #03/16\n|[![Star](https://img.shields.io/github/stars/da03/implicit_chain_of_thought.svg?style=social\u0026label=Star)](https://github.com/da03/implicit_chain_of_thought/)\u003cbr\u003e[Implicit Chain of Thought Reasoning via Knowledge Distillation](https://arxiv.org/abs/2311.01460) \u003cbr\u003e Yuntian Deng, Kiran Prasad, Roland Fernandez, Paul Smolensky, Vishrav Chaudhary, Stuart Shieber |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/explicit2implicit.png\"\u003e |[Github](https://github.com/da03/implicit_chain_of_thought/) \u003cbr\u003e [Paper](https://arxiv.org/abs/2311.01460)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/HKUNLP/diffusion-of-thoughts.svg?style=social\u0026label=Star)](https://github.com/HKUNLP/diffusion-of-thoughts) [![Publish](https://img.shields.io/badge/Conference-NeurIPS_2024-blue)]()\u003cbr\u003e[Diffusion of Thoughts: Chain-of-Thought Reasoning in Diffusion Language Models](https://arxiv.org/abs/2402.07754) \u003cbr\u003e Jiacheng Ye, Shansan Gong, Liheng Chen, Lin Zheng, Jiahui Gao, Han Shi, Chuan Wu, Xin Jiang, Zhenguo Li, Wei Bi, Lingpeng Kong |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/diffusion_thought.png\"\u003e |[Github](https://github.com/HKUNLP/diffusion-of-thoughts) \u003cbr\u003e [Paper](https://arxiv.org/abs/2402.07754)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/da03/Internalize_CoT_Step_by_Step.svg?style=social\u0026label=Star)](https://github.com/da03/Internalize_CoT_Step_by_Step)\u003cbr\u003e[From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step](https://arxiv.org/abs/2405.14838) \u003cbr\u003e Yuntian Deng, Yejin Choi, Stuart Shieber |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2405.14838v1/extracted/2405.14838v1/training_illustration.png\"\u003e |[Github](https://github.com/da03/Internalize_CoT_Step_by_Step) \u003cbr\u003e [Paper](https://arxiv.org/abs/2405.14838)| [//]: #04/08\n|[Compressed Chain of Thought: Efficient Reasoning Through Dense Representations](https://arxiv.org/abs/2412.13171) \u003cbr\u003e Jeffrey Cheng, Benjamin Van Durme |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2412.13171v1/extracted/6074157/figures/fig1.png\"\u003e |[Paper](https://arxiv.org/abs/2412.13171)| [//]: #04/08\n|[SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs](https://arxiv.org/abs/2502.12134) \u003cbr\u003e Yige Xu, Xu Guo, Zhiwei Zeng, Chunyan Miao |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.12134v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2502.12134)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-ICLR_2025-blue)]()\u003cbr\u003e[Reasoning with Latent Thoughts: On the Power of Looped Transformers](https://arxiv.org/abs/2502.17416) \u003cbr\u003e Nikunj Saunshi, Nishanth Dikkala, Zhiyuan Li, Sanjiv Kumar, Sashank J. Reddi |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.17416v1/extracted/6229618/Media/looping_illustration2.png\"\u003e |[Paper](https://arxiv.org/abs/2502.17416)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/qifanyu/RELAY.svg?style=social\u0026label=Star)](https://github.com/qifanyu/RELAY)\u003cbr\u003e[Enhancing Auto-regressive Chain-of-Thought through Loop-Aligned Reasoning](https://arxiv.org/abs/2502.08482) \u003cbr\u003e Qifan Yu, Zhenyu He, Sijie Li, Xun Zhou, Jun Zhang, Jingjing Xu, Di He |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.08482v1/x1.png\"\u003e |[Github](https://github.com/qifanyu/RELAY) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.08482)| [//]: #04/08\n|[CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation](https://arxiv.org/abs/2502.21074) \u003cbr\u003e Zhenyi Shen, Hanqi Yan, Linhai Zhang, Zhanghao Hu, Yali Du, Yulan He |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.21074v1/extracted/6241542/figures/codi_illustrate12.png\"\u003e |[Paper](https://arxiv.org/abs/2502.21074)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/zjunlp/LightThinker.svg?style=social\u0026label=Star)](https://github.com/zjunlp/LightThinker)\u003cbr\u003e[LightThinker: Thinking Step-by-Step Compression](https://arxiv.org/abs/2502.15589) \u003cbr\u003e Jintian Zhang, Yuqi Zhu, Mengshu Sun, Yujie Luo, Shuofei Qiao, Lun Du, Da Zheng, Huajun Chen, Ningyu Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.15589v1/x1.png\"\u003e |[Github](https://github.com/zjunlp/LightThinker) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.15589)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/WANGXinyiLinda/planning_tokens.svg?style=social\u0026label=Star)](https://github.com/WANGXinyiLinda/planning_tokens) [![Publish](https://img.shields.io/badge/Conference-COLM_2024-blue)]()\u003cbr\u003e[Guiding Language Model Reasoning with Planning Tokens](https://arxiv.org/abs/2310.05707) \u003cbr\u003e Xinyi Wang, Lucas Caccia, Oleksiy Ostapenko, Xingdi Yuan, William Yang Wang, Alessandro Sordoni |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2310.05707v4/extracted/5777851/img/overview.png\"\u003e |[Github](https://github.com/WANGXinyiLinda/planning_tokens) \u003cbr\u003e [Paper](https://arxiv.org/abs/2310.05707)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/JacobPfau/fillerTokens.svg?style=social\u0026label=Star)](https://github.com/JacobPfau/fillerTokens) [![Publish](https://img.shields.io/badge/Conference-COLM_2024-blue)]()\u003cbr\u003e[Let's Think Dot by Dot: Hidden Computation in Transformer Language Models](https://arxiv.org/abs/2404.15758) \u003cbr\u003e Jacob Pfau, William Merrill, Samuel R. Bowman |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2404.15758v1/extracted/2404.15758v1/figs/scale_len.png\"\u003e |[Github](https://github.com/JacobPfau/fillerTokens) \u003cbr\u003e [Paper](https://arxiv.org/abs/2404.15758)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/MingyuJ666/Disentangling-Memory-and-Reasoning.svg?style=social\u0026label=Star)](https://github.com/MingyuJ666/Disentangling-Memory-and-Reasoning)\u003cbr\u003e[Disentangling Memory and Reasoning Ability in Large Language Models](https://arxiv.org/abs/2411.13504) \u003cbr\u003e Mingyu Jin, Weidi Luo, Sitao Cheng, Xinyi Wang, Wenyue Hua, Ruixiang Tang, William Yang Wang, Yongfeng Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2411.13504v2/x1.png\"\u003e |[Github](https://github.com/MingyuJ666/Disentangling-Memory-and-Reasoning) \u003cbr\u003e [Paper](https://arxiv.org/abs/2411.13504)| [//]: #04/08\n|[Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning](https://arxiv.org/abs/2502.03275) \u003cbr\u003e DiJia Su, Hanlin Zhu, Yingchen Xu, Jiantao Jiao, Yuandong Tian, Qinqing Zheng |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.03275v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2502.03275)| [//]: #04/08\n|[Training Large Language Models to Reason in a Continuous Latent Space](https://arxiv.org/abs/2412.06769) \u003cbr\u003e Shibo Hao, Sainbayar Sukhbaatar, DiJia Su, Xian Li, Zhiting Hu, Jason Weston, Yuandong Tian |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2412.06769v2/extracted/6060815/figures/figure_1_meta_3.png\"\u003e |[Paper](https://arxiv.org/abs/2412.06769)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/shawnricecake/Heima.svg?style=social\u0026label=Star)](https://github.com/shawnricecake/Heima)\u003cbr\u003e[Efficient Reasoning with Hidden Thinking](https://arxiv.org/abs/2501.19201) \u003cbr\u003e Xuan Shen, Yizhou Wang, Xiangxi Shi, Yanzhi Wang, Pu Zhao, Jiuxiang Gu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2501.19201v1/x1.png\"\u003e |[Github](https://github.com/shawnricecake/Heima) \u003cbr\u003e [Paper](https://arxiv.org/abs/2501.19201)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-ICLR_2024-blue)]()\u003cbr\u003e[Think before you speak: Training Language Models With Pause Tokens](https://arxiv.org/abs/2310.02226) \u003cbr\u003e Sachin Goyal, Ziwei Ji, Ankit Singh Rawat, Aditya Krishna Menon, Sanjiv Kumar, Vaishnavh Nagarajan |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/pause_token.png\"\u003e |[Paper](https://arxiv.org/abs/2310.02226)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/seal-rg/recurrent-pretraining.svg?style=social\u0026label=Star)](https://github.com/seal-rg/recurrent-pretraining)\u003cbr\u003e[Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach](https://arxiv.org/abs/2502.05171) \u003cbr\u003e Jonas Geiping, Sean McLeish, Neel Jain, John Kirchenbauer, Siddharth Singh, Brian R. Bartoldson, Bhavya Kailkhura, Abhinav Bhatele, Tom Goldstein |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.05171v2/x2.png\"\u003e |[Github](https://github.com/seal-rg/recurrent-pretraining) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.05171)| [//]: #04/08\n|[Weight-of-Thought Reasoning: Exploring Neural Network Weights for Enhanced LLM Reasoning](https://arxiv.org/abs/2504.10646) \u003cbr\u003e Saif Punjwani, Larry Heck |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.10646v1/extracted/6355099/cotvswot.png\"\u003e |[Paper](https://arxiv.org/abs/2504.10646)|[//]: #04/16\n\n\n### Build SLM with Strong Reasoning Ability\n\n\n#### Distillation \n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n| [![Publish](https://img.shields.io/badge/Conference-ACL_2023-blue)]()\u003cbr\u003e[Teaching Small Language Models to Reason](https://arxiv.org/abs/2212.08410) \u003cbr\u003e Lucie Charlotte Magister, Jonathan Mallinson, Jakub Adamek, Eric Malmi, Aliaksei Severyn |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/slm_kd.png\"\u003e |[Paper](https://arxiv.org/abs/2212.08410)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-EMNLP_2024-blue)]()\u003cbr\u003e[Mixed Distillation Helps Smaller Language Model Better Reasoning](https://arxiv.org/abs/2312.10730) \u003cbr\u003e Chenglin Li, Qianglong Chen, Liangyue Li, Caiyu Wang, Yicheng Li, Zulong Chen, Yin Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/mix_distillation.png\"\u003e |[Paper](https://arxiv.org/abs/2312.10730)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Small-Model-Gap/Small-Model-Learnability-Gap.svg?style=social\u0026label=Star)](https://github.com/Small-Model-Gap/Small-Model-Learnability-Gap)\u003cbr\u003e[Small Models Struggle to Learn from Strong Reasoners](https://arxiv.org/abs/2502.12143) \u003cbr\u003e Yuetai Li, Xiang Yue, Zhangchen Xu, Fengqing Jiang, Luyao Niu, Bill Yuchen Lin, Bhaskar Ramasubramanian, Radha Poovendran |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.12143v2/x1.png\"\u003e |[Github](https://github.com/Small-Model-Gap/Small-Model-Learnability-Gap) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.12143)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Yiwei98/TDG.svg?style=social\u0026label=Star)](https://github.com/Yiwei98/TDG) [![Publish](https://img.shields.io/badge/Conference-AAAI_2024-blue)]()\u003cbr\u003e[Turning Dust into Gold: Distilling Complex Reasoning Capabilities from LLMs by Leveraging Negative Data](https://arxiv.org/abs/2312.12832) \u003cbr\u003e Yiwei Li, Peiwen Yuan, Shaoxiong Feng, Boyuan Pan, Bin Sun, Xinglin Wang, Heda Wang, Kan Li |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2312.12832v1/x1.png\"\u003e |[Github](https://github.com/Yiwei98/TDG) \u003cbr\u003e [Paper](https://arxiv.org/abs/2312.12832)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-EMNLP_2024-blue)]()\u003cbr\u003e[Teaching Small Language Models Reasoning through Counterfactual Distillation](https://aclanthology.org/2024.emnlp-main.333/) \u003cbr\u003e Tao Feng, Yicheng Li, Li Chenglin, Hao Chen, Fei Yu, Yin Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/counterfactual_distillation.png\"\u003e |[Paper](https://aclanthology.org/2024.emnlp-main.333/)| [//]: #04/08\n|[Deconstructing Long Chain-of-Thought: A Structured Reasoning Optimization Framework for Long CoT Distillation](https://arxiv.org/abs/2503.16385) \u003cbr\u003e Yijia Luo, Yulin Song, Xingyao Zhang, Jiaheng Liu, Weixun Wang, GengRu Chen, Wenbo Su, Bo Zheng |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.16385v1/x3.png\"\u003e |[Paper](https://arxiv.org/abs/2503.16385)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/yunx-z/SCORE.svg?style=social\u0026label=Star)](https://github.com/yunx-z/SCORE) [![Publish](https://img.shields.io/badge/Conference-ACL_Findings_2024-blue)]()\u003cbr\u003e[Small Language Models Need Strong Verifiers to Self-Correct Reasoning](https://arxiv.org/abs/2404.17140) \u003cbr\u003e Yunxiang Zhang, Muhammad Khalifa, Lajanugen Logeswaran, Jaekyeom Kim, Moontae Lee, Honglak Lee, Lu Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2404.17140v2/x1.png\"\u003e |[Github](https://github.com/yunx-z/SCORE) \u003cbr\u003e [Paper](https://arxiv.org/abs/2404.17140)| [//]: #04/08\n|[Improving Mathematical Reasoning Capabilities of Small Language Models via Feedback-Driven Distillation](https://arxiv.org/abs/2411.14698) \u003cbr\u003e Xunyu Zhu, Jian Li, Can Ma, Weiping Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2411.14698v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2411.14698)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Xnhyacinth/SKIntern.svg?style=social\u0026label=Star)](https://github.com/Xnhyacinth/SKIntern) [![Publish](https://img.shields.io/badge/Conference-COLING_2025-blue)]()\u003cbr\u003e[SKIntern : Internalizing Symbolic Knowledge for Distilling Better CoT Capabilities into Small Language Models](https://arxiv.org/abs/2409.13183) \u003cbr\u003e Huanxuan Liao, Shizhu He, Yupu Hao, Xiang Li, Yuanzhe Zhang, Jun Zhao, Kang Liu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2409.13183v2/x1.png\"\u003e |[Github](https://github.com/Xnhyacinth/SKIntern) \u003cbr\u003e [Paper](https://arxiv.org/abs/2409.13183)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-COLING_2024-blue)]()\u003cbr\u003e[Probe then Retrieve and Reason: Distilling Probing and Reasoning Capabilities into Smaller Language Models](https://aclanthology.org/2024.lrec-main.1140.pdf) \u003cbr\u003e Yichun Zhao, Shuheng Zhou, Huijia Zhu |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/prr.png\"\u003e |[Paper](https://aclanthology.org/2024.lrec-main.1140.pdf)| [//]: #04/08\n|[Thinking Slow, Fast: Scaling Inference Compute with Distilled Reasoners](https://arxiv.org/abs/2502.20339) \u003cbr\u003e Daniele Paliotta, Junxiong Wang, Matteo Pagliardini, Kevin Y. Li, Aviv Bick, J. Zico Kolter, Albert Gu, François Fleuret, Tri Dao |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.20339v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2502.20339)| [//]: #04/08\n|[Distilling Reasoning Ability from Large Language Models with Adaptive Thinking](https://arxiv.org/abs/2404.09170) \u003cbr\u003e Xiaoshu Chen, Sihang Zhou, Ke Liang, Xinwang Liu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2404.09170v5/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2404.09170)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/EIT-NLP/Distilling-CoT-Reasoning.svg?style=social\u0026label=Star)](https://github.com/EIT-NLP/Distilling-CoT-Reasoning)\u003cbr\u003e[Unveiling the Key Factors for Distilling Chain-of-Thought Reasoning](https://arxiv.org/abs/2502.18001) \u003cbr\u003e Xinghao Chen, Zhijing Sun, Wenjin Guo, Miaoran Zhang, Yanjun Chen, Yirong Sun, Hui Su, Yijie Pan, Dietrich Klakow, Wenjie Li, Xiaoyu Shen |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.18001v1/x1.png\"\u003e |[Github](https://github.com/EIT-NLP/Distilling-CoT-Reasoning) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.18001)| [//]: #04/08\n\n#### Quantization and Pruning\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[Towards Reasoning Ability of Small Language Models](https://arxiv.org/abs/2502.11569) \u003cbr\u003e Gaurav Srivastava, Shuxiang Cao, Xuan Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/slm_reasoning.png\"\u003e |[Paper](https://arxiv.org/abs/2502.11569)| [//]: #04/14\n|[![Star](https://img.shields.io/github/stars/ruikangliu/Quantized-Reasoning-Models.svg?style=social\u0026label=Star)](https://github.com/ruikangliu/Quantized-Reasoning-Models)\u003cbr\u003e[Quantization Hurts Reasoning? An Empirical Study on Quantized Reasoning Models](https://arxiv.org/abs/2504.04823) \u003cbr\u003e Ruikang Liu, Yuxuan Sun, Manyi Zhang, Haoli Bai, Xianzhi Yu, Tiezheng Yu, Chun Yuan, Lu Hou |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/quant_hurt.png\"\u003e |[Github](https://github.com/ruikangliu/Quantized-Reasoning-Models) \u003cbr\u003e [Paper](https://arxiv.org/abs/2504.04823)| [//]: #04/14\n|[When Reasoning Meets Compression: Benchmarking Compressed Large Reasoning Models on Complex Reasoning Tasks](https://arxiv.org/abs/2504.02010) \u003cbr\u003e Nan Zhang, Yusen Zhang, Prasenjit Mitra, Rui Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/when_compression.png\"\u003e |[Paper](https://arxiv.org/abs/2504.02010)| [//]: #04/14\n\n\n\n#### RL-based Methods\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[![Star](https://img.shields.io/github/stars/knoveleng/open-rs.svg?style=social\u0026label=Star)](https://github.com/knoveleng/open-rs)\u003cbr\u003e[Reinforcement Learning for Reasoning in Small LLMs: What Works and What Doesn't](https://arxiv.org/abs/2503.16219) \u003cbr\u003e Quy-Anh Dang, Chris Ngo |\u003cimg src=\"https://arxiv.org/html/2503.16219v1/extracted/6296504/images/pass1.png\" width=\"45%\"\u003e \u003cimg src=\"https://arxiv.org/html/2503.16219v1/extracted/6296504/images/costs.png\" width=\"45%\"\u003e |[Github](https://github.com/knoveleng/open-rs) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.16219)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/hkust-nlp/simpleRL-reason.svg?style=social\u0026label=Star)](https://github.com/hkust-nlp/simpleRL-reason)\u003cbr\u003e[SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild](https://arxiv.org/abs/2503.18892) \u003cbr\u003e Weihao Zeng, Yuzhen Huang, Qian Liu, Wei Liu, Keqing He, Zejun Ma, Junxian He |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/simplerl_zoo.png\"\u003e |[Github](https://github.com/hkust-nlp/simpleRL-reason) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.18892)| [//]: #04/08\n\n###### Repo\n\n* [DeepScaleR](https://github.com/agentica-project/deepscaler). DeepScaleR team. [Webpage](https://agentica-project.com/)\n\n\n\n### Let Decoding More Efficient\n\n\n#### Efficient TTS\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[Think Deep, Think Fast: Investigating Efficiency of Verifier-free Inference-time-scaling Methods](https://arxiv.org/abs/2504.14047) \u003cbr\u003e Junlin Wang, Shang Zhu, Jon Saad-Falcon, Ben Athiwaratkun, Qingyang Wu, Jue Wang, Shuaiwen Leon Song, Ce Zhang, Bhuwan Dhingra, James Zou |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.14047v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2504.14047)| [//]: #04/23\n|[![Star](https://img.shields.io/github/stars/IAAR-Shanghai/xVerify.svg?style=social\u0026label=Star)](https://github.com/IAAR-Shanghai/xVerify)\u003cbr\u003e[xVerify: Efficient Answer Verifier for Reasoning Model Evaluations](https://arxiv.org/abs/2504.10481) \u003cbr\u003e Ding Chen, Qingchen Yu, Pengyuan Wang, Wentao Zhang, Bo Tang, Feiyu Xiong, Xinchi Li, Minchuan Yang, Zhiyu Li |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.10481v1/x1.png\"\u003e |[Github](https://github.com/IAAR-Shanghai/xVerify) \u003cbr\u003e [Paper](https://arxiv.org/abs/2504.10481)| [//]: #04/17\n|[![Star](https://img.shields.io/github/stars/Pranjal2041/AdaptiveConsistency.svg?style=social\u0026label=Star)](https://github.com/Pranjal2041/AdaptiveConsistency)\u003cbr\u003e[Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMs](https://arxiv.org/abs/2305.11860) \u003cbr\u003e Pranjal Aggarwal, Aman Madaan, Yiming Yang, Mausam |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/asc.png\"\u003e |[Github](https://github.com/Pranjal2041/AdaptiveConsistency) \u003cbr\u003e [Paper](https://arxiv.org/abs/2305.11860)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Yiwei98/ESC.svg?style=social\u0026label=Star)](https://github.com/Yiwei98/ESC) [![Publish](https://img.shields.io/badge/Conference-ICLR_2024-blue)]()\u003cbr\u003e[Escape Sky-high Cost: Early-stopping Self-Consistency for Multi-step Reasoning](https://arxiv.org/abs/2401.10480) \u003cbr\u003e Yiwei Li, Peiwen Yuan, Shaoxiong Feng, Boyuan Pan, Xinglin Wang, Bin Sun, Heda Wang, Kan Li |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2401.10480v1/x1.png\"\u003e |[Github](https://github.com/Yiwei98/ESC) \u003cbr\u003e [Paper](https://arxiv.org/abs/2401.10480)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/WangXinglin/DSC.svg?style=social\u0026label=Star)](https://github.com/WangXinglin/DSC) [![Publish](https://img.shields.io/badge/Conference-NAACL_Findings_2025-blue)]()\u003cbr\u003e[Make Every Penny Count: Difficulty-Adaptive Self-Consistency for Cost-Efficient Reasoning](https://arxiv.org/abs/2408.13457) \u003cbr\u003e Xinglin Wang, Shaoxiong Feng, Yiwei Li, Peiwen Yuan, Yueqi Zhang, Chuyi Tan, Boyuan Pan, Yao Hu, Kan Li |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2408.13457v3/x3.png\"\u003e |[Github](https://github.com/WangXinglin/DSC) \u003cbr\u003e [Paper](https://arxiv.org/abs/2408.13457)| [//]: #04/08\n|[Path-Consistency: Prefix Enhancement for Efficient Inference in LLM](https://arxiv.org/abs/2409.01281) \u003cbr\u003e Jiace Zhu, Yingtao Shen, Jie Zhao, An Zou |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2409.01281v2/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2409.01281)| [//]: #04/08\n|[Bridging Internal Probability and Self-Consistency for Effective and Efficient LLM Reasoning](https://arxiv.org/abs/2502.00511) \u003cbr\u003e Zhi Zhou, Tan Yuhao, Zenan Li, Yuan Yao, Lan-Zhe Guo, Xiaoxing Ma, Yu-Feng Li |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.00511v2/x3.png\"\u003e |[Paper](https://arxiv.org/abs/2502.00511)| [//]: #04/08\n|[Confidence Improves Self-Consistency in LLMs](https://arxiv.org/abs/2502.06233) \u003cbr\u003e Amir Taubenfeld, Tom Sheffer, Eran Ofek, Amir Feder, Ariel Goldstein, Zorik Gekhman, Gal Yona |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/cisc.png\"\u003e |[Paper](https://arxiv.org/abs/2502.06233)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Chengsong-Huang/Self-Calibration.svg?style=social\u0026label=Star)](https://github.com/Chengsong-Huang/Self-Calibration)\u003cbr\u003e[Efficient Test-Time Scaling via Self-Calibration](https://arxiv.org/abs/2503.00031) \u003cbr\u003e Chengsong Huang, Langlin Huang, Jixuan Leng, Jiacheng Liu, Jiaxin Huang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.00031v1/x2.png\"\u003e |[Github](https://github.com/Chengsong-Huang/Self-Calibration) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.00031)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Zanette-Labs/SpeculativeRejection.svg?style=social\u0026label=Star)](https://github.com/Zanette-Labs/SpeculativeRejection) [![Publish](https://img.shields.io/badge/Conference-NeurIPS_2024-blue)]()\u003cbr\u003e[Fast Best-of-N Decoding via Speculative Rejection](https://arxiv.org/abs/2410.20290) \u003cbr\u003e Hanshi Sun, Momin Haider, Ruiqi Zhang, Huitao Yang, Jiahao Qiu, Ming Yin, Mengdi Wang, Peter Bartlett, Andrea Zanette |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2410.20290v2/x1.png\"\u003e |[Github](https://github.com/Zanette-Labs/SpeculativeRejection) \u003cbr\u003e [Paper](https://arxiv.org/abs/2410.20290)| [//]: #04/08\n|[Sampling-Efficient Test-Time Scaling: Self-Estimating the Best-of-N Sampling in Early Decoding](https://arxiv.org/abs/2503.01422) \u003cbr\u003e Yiming Wang, Pei Zhang, Siyuan Huang, Baosong Yang, Zhuosheng Zhang, Fei Huang, Rui Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.01422v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2503.01422)| [//]: #04/08\n|[FastMCTS: A Simple Sampling Strategy for Data Synthesis](https://www.arxiv.org/abs/2502.11476) \u003cbr\u003e Peiji Li, Kai Lv, Yunfan Shao, Yichuan Ma, Linyang Li, Xiaoqing Zheng, Xipeng Qiu, Qipeng Guo |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.11476v1/x2.png\"\u003e |[Paper](https://www.arxiv.org/abs/2502.11476)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/chang-github-00/LLM-Predictive-Decoding.svg?style=social\u0026label=Star)](https://github.com/chang-github-00/LLM-Predictive-Decoding) [![Publish](https://img.shields.io/badge/Conference-ICLR_2025-blue)]()\u003cbr\u003e[Non-myopic Generation of Language Models for Reasoning and Planning](https://arxiv.org/abs/2410.17195) \u003cbr\u003e Chang Ma, Haiteng Zhao, Junlei Zhang, Junxian He, Lingpeng Kong |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/predictive_decoding.png\"\u003e |[Github](https://github.com/chang-github-00/LLM-Predictive-Decoding) \u003cbr\u003e [Paper](https://arxiv.org/abs/2410.17195)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/ethanm88/self-taught-lookahead.svg?style=social\u0026label=Star)](https://github.com/ethanm88/self-taught-lookahead)\u003cbr\u003e[Language Models can Self-Improve at State-Value Estimation for Better Search](https://arxiv.org/abs/2503.02878) \u003cbr\u003e Ethan Mendes, Alan Ritter |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.02878v1/x1.png\"\u003e |[Github](https://github.com/ethanm88/self-taught-lookahead) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.02878)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/xufangzhi/phi-Decoding.svg?style=social\u0026label=Star)](https://github.com/xufangzhi/phi-Decoding)\u003cbr\u003e[ϕ-Decoding: Adaptive Foresight Sampling for Balanced Inference-Time Exploration and Exploitation](https://arxiv.org/abs/2503.13288) \u003cbr\u003e Fangzhi Xu, Hang Yan, Chang Ma, Haiteng Zhao, Jun Liu, Qika Lin, Zhiyong Wu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.13288v1/x2.png\"\u003e |[Github](https://github.com/xufangzhi/phi-Decoding) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.13288)| [//]: #04/08\n|[Dynamic Parallel Tree Search for Efficient LLM Reasoning](https://arxiv.org/abs/2502.16235) \u003cbr\u003e Yifu Ding, Wentao Jiang, Shunyu Liu, Yongcheng Jing, Jinyang Guo, Yingjie Wang, Jing Zhang, Zengmao Wang, Ziwei Liu, Bo Du, Xianglong Liu, Dacheng Tao |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.16235v2/x5.png\"\u003e |[Paper](https://arxiv.org/abs/2502.16235)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Soistesimmer/Fetch.svg?style=social\u0026label=Star)](https://github.com/Soistesimmer/Fetch)\u003cbr\u003e[Don't Get Lost in the Trees: Streamlining LLM Reasoning by Overcoming Tree Search Exploration Pitfalls](https://arxiv.org/abs/2502.11183) \u003cbr\u003e Ante Wang, Linfeng Song, Ye Tian, Dian Yu, Haitao Mi, Xiangyu Duan, Zhaopeng Tu, Jinsong Su, Dong Yu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.11183v2/extracted/6301324/figures/method.png\"\u003e |[Github](https://github.com/Soistesimmer/Fetch) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.11183)| [//]: #04/08\n\n\n\n#### Other Optimal Methods\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[![Star](https://img.shields.io/github/stars/Parallel-Reasoning/APR.svg?style=social\u0026label=Star)](https://github.com/Parallel-Reasoning/APR)\u003cbr\u003e[Learning Adaptive Parallel Reasoning with Language Models](https://arxiv.org/abs/2504.15466) \u003cbr\u003e Jiayi Pan, Xiuyu Li, Long Lian, Charlie Snell, Yifei Zhou, Adam Yala, Trevor Darrell, Kurt Keutzer, Alane Suhr |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.15466v1/x2.png\"\u003e |[Github](https://github.com/Parallel-Reasoning/APR) \u003cbr\u003e [Paper](https://arxiv.org/abs/2504.15466)| [//]: #04/23\n|[THOUGHTTERMINATOR: Benchmarking, Calibrating, and Mitigating Overthinking in Reasoning Models](https://arxiv.org/abs/2504.13367) \u003cbr\u003e Xiao Pu, Michael Saxon, Wenyue Hua, William Yang Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.13367v1/x2.png\"\u003e |[Paper](https://arxiv.org/abs/2504.13367)| [//]: #04/21\n|[![Star](https://img.shields.io/github/stars/imagination-research/sot.svg?style=social\u0026label=Star)](https://github.com/imagination-research/sot) [![Publish](https://img.shields.io/badge/Conference-ICLR_2024-blue)]()\u003cbr\u003e[Skeleton-of-Thought: Prompting LLMs for Efficient Parallel Generation](https://arxiv.org/abs/2307.15337) \u003cbr\u003e Xuefei Ning, Zinan Lin, Zixuan Zhou, Zifu Wang, Huazhong Yang, Yu Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/skeleton_ot.png\"\u003e |[Github](https://github.com/imagination-research/sot) \u003cbr\u003e [Paper](https://arxiv.org/abs/2307.15337)| [//]: #04/08\n|[Adaptive Skeleton Graph Decoding](https://arxiv.org/abs/2402.12280) \u003cbr\u003e Shuowei Jin, Yongji Wu, Haizhong Zheng, Qingzhao Zhang, Matthew Lentz, Z. Morley Mao, Atul Prakash, Feng Qian, Danyang Zhuo |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/sgd.png\"\u003e |[Paper](https://arxiv.org/abs/2402.12280)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/BaohaoLiao/RSD.svg?style=social\u0026label=Star)](https://github.com/BaohaoLiao/RSD)\u003cbr\u003e[Reward-Guided Speculative Decoding for Efficient LLM Reasoning](https://arxiv.org/abs/2501.19324) \u003cbr\u003e Baohao Liao, Yuhui Xu, Hanze Dong, Junnan Li, Christof Monz, Silvio Savarese, Doyen Sahoo, Caiming Xiong |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/rsd.png\"\u003e |[Github](https://github.com/BaohaoLiao/RSD) \u003cbr\u003e [Paper](https://arxiv.org/abs/2501.19324)| [//]: #04/08\n|[Meta-Reasoner: Dynamic Guidance for Optimized Inference-time Reasoning in Large Language Models](https://arxiv.org/abs/2502.19918) \u003cbr\u003e Yuan Sui, Yufei He, Tri Cao, Simeng Han, Bryan Hooi |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/meta_reasoner.png\"\u003e |[Paper](https://arxiv.org/abs/2502.19918)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/qixucen/atom.svg?style=social\u0026label=Star)](https://github.com/qixucen/atom)\u003cbr\u003e[Atom of Thoughts for Markov LLM Test-Time Scaling](https://arxiv.org/abs/2502.12018) \u003cbr\u003e Fengwei Teng, Zhaoyang Yu, Quan Shi, Jiayi Zhang, Chenglin Wu, Yuyu Luo |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/aot.png\"\u003e |[Github](https://github.com/qixucen/atom) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.12018)| [//]: #04/08\n|[DISC: Dynamic Decomposition Improves LLM Inference Scaling](https://arxiv.org/abs/2502.16706) \u003cbr\u003e Jonathan Light, Wei Cheng, Wu Yue, Masafumi Oyamada, Mengdi Wang, Santiago Paternain, Haifeng Chen |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/disc.png\"\u003e |[Paper](https://arxiv.org/abs/2502.16706)| [//]: #04/08\n|[From Chaos to Order: The Atomic Reasoner Framework for Fine-grained Reasoning in Large Language Models](https://arxiv.org/abs/2503.15944) \u003cbr\u003e Jinyi Liu, Yan Zheng, Rong Cheng, Qiyu Wu, Wei Guo, Fei Ni, Hebin Liang, Yifu Yuan, Hangyu Mao, Fuzheng Zhang, Jianye Hao |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/ar.png\"\u003e |[Paper](https://arxiv.org/abs/2503.15944)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/Quinn777/AtomThink.svg?style=social\u0026label=Star)](https://github.com/Quinn777/AtomThink)\u003cbr\u003e[Can Atomic Step Decomposition Enhance the Self-structured Reasoning of Multimodal Large Models?](https://arxiv.org/abs/2503.06252) \u003cbr\u003e Kun Xiang, Zhili Liu, Zihao Jiang, Yunshuang Nie, Kaixin Cai, Yiyang Yin, Runhui Huang, Haoxiang Fan, Hanhui Li, Weiran Huang, Yihan Zeng, Yu-Jie Yuan, Jianhua Han, Lanqing Hong, Hang Xu, Xiaodan Liang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/atom.png\"\u003e |[Github](https://github.com/Quinn777/AtomThink) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.06252)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-ICLR_2025-blue)]()\u003cbr\u003e[Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://arxiv.org/abs/2408.03314) \u003cbr\u003e Charlie Snell, Jaehoon Lee, Kelvin Xu, Aviral Kumar |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/tts_effective.png\"\u003e |[Paper](https://arxiv.org/abs/2408.03314)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/thu-wyz/inference_scaling.svg?style=social\u0026label=Star)](https://github.com/thu-wyz/inference_scaling) [![Publish](https://img.shields.io/badge/Conference-ICLR_2025-blue)]()\u003cbr\u003e[Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models](https://arxiv.org/abs/2408.00724) \u003cbr\u003e Yangzhen Wu, Zhiqing Sun, Shanda Li, Sean Welleck, Yiming Yang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/scaling_law.png\"\u003e |[Github](https://github.com/thu-wyz/inference_scaling) \u003cbr\u003e [Paper](https://arxiv.org/abs/2408.00724)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/CMU-AIRe/MRT.svg?style=social\u0026label=Star)](https://github.com/CMU-AIRe/MRT)\u003cbr\u003e[Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuning](https://arxiv.org/abs/2503.07572) \u003cbr\u003e Yuxiao Qu, Matthew Y. R. Yang, Amrith Setlur, Lewis Tunstall, Edward Emanuel Beeching, Ruslan Salakhutdinov, Aviral Kumar |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/mrt.png\"\u003e |[Github](https://github.com/CMU-AIRe/MRT) \u003cbr\u003e [Paper](https://arxiv.org/abs/2503.07572)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/ruipeterpan/specreason.svg?style=social\u0026label=Star)](https://github.com/ruipeterpan/specreason)\u003cbr\u003e[SpecReason: Fast and Accurate Inference-Time Compute via Speculative Reasoning](https://arxiv.org/abs/2504.07891) \u003cbr\u003e Rui Pan, Yinwei Dai, Zhihao Zhang, Gabriele Oliaro, Zhihao Jia, Ravi Netravali |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/specreason.png\"\u003e |[Github](https://github.com/ruipeterpan/specreason) \u003cbr\u003e [Paper](https://arxiv.org/abs/2504.07891)| [//]: #04/14\n\n### Evaluation and Benchmarks\n\n\n#### Metric\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs](https://arxiv.org/abs/2412.21187) \u003cbr\u003e 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, Dong Yu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2412.21187v2/x2.png\"\u003e |[Paper](https://arxiv.org/abs/2412.21187)|[//]: #03/16\n|[![Star](https://img.shields.io/github/stars/horseee/CoT-Valve.svg?style=social\u0026label=Star)](https://github.com/horseee/CoT-Valve)\u003cbr\u003e[CoT-Valve: Length-Compressible Chain-of-Thought Tuning](https://arxiv.org/abs/2502.09601) \u003cbr\u003e Xinyin Ma, Guangnian Wan, Runpeng Yu, Gongfan Fang, Xinchao Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/cot_valve.png\"\u003e |[Github](https://github.com/horseee/CoT-Valve) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.09601)|[//]: #03/16\n|[![Star](https://img.shields.io/github/stars/breckbaldwin/llm-stability.svg?style=social\u0026label=Star)](https://github.com/breckbaldwin/llm-stability)\u003cbr\u003e[Non-Determinism of \"Deterministic\" LLM Settings](https://arxiv.org/abs/2408.04667) \u003cbr\u003e Berk Atil, Sarp Aykent, Alexa Chittams, Lisheng Fu, Rebecca J. Passonneau, Evan Radcliffe, Guru Rajan Rajagopal, Adam Sloan, Tomasz Tudrej, Ferhan Ture, Zhe Wu, Lixinyu Xu, Breck Baldwin |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2408.04667v5/extracted/6331111/max_min_diff.png\"\u003e |[Github](https://github.com/breckbaldwin/llm-stability) \u003cbr\u003e [Paper](https://arxiv.org/abs/2408.04667)| [//]: #04/08\n|[The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks](https://arxiv.org/abs/2502.08235) \u003cbr\u003e Alejandro Cuadron, Dacheng Li, Wenjie Ma, Xingyao Wang, Yichuan Wang, Siyuan Zhuang, Shu Liu, Luis Gaspar Schroeder, Tian Xia, Huanzhi Mao, Nicholas Thumiger, Aditya Desai, Ion Stoica, Ana Klimovic, Graham Neubig, Joseph E. Gonzalez |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/openhands.png\"\u003e |[Paper](https://arxiv.org/abs/2502.08235)| [//]: #04/08\n|[Evaluating Large Language Models Trained on Code](https://arxiv.org/abs/2107.03374) \u003cbr\u003e Mark Chen, Jerry Tworek, et al. |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/passk.png\"\u003e |[Paper](https://arxiv.org/abs/2107.03374)| [//]: #04/08\n|[τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://arxiv.org/abs/2406.12045) \u003cbr\u003e Shunyu Yao, Noah Shinn, Pedram Razavi, Karthik Narasimhan |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/agent_bench.png\"\u003e |[Paper](https://arxiv.org/abs/2406.12045)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/open-compass/GPassK.svg?style=social\u0026label=Star)](https://github.com/open-compass/GPassK)\u003cbr\u003e[Are Your LLMs Capable of Stable Reasoning?](https://arxiv.org/abs/2412.13147) \u003cbr\u003e Junnan Liu, Hongwei Liu, Linchen Xiao, Ziyi Wang, Kuikun Liu, Songyang Gao, Wenwei Zhang, Songyang Zhang, Kai Chen |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2412.13147v3/x1.png\"\u003e |[Github](https://github.com/open-compass/GPassK) \u003cbr\u003e [Paper](https://arxiv.org/abs/2412.13147)| [//]: #04/08\n\n\n#### Benchmarks and Datasets\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[LongPerceptualThoughts: Distilling System-2 Reasoning for System-1 Perception](https://arxiv.org/abs/2504.15362) \u003cbr\u003e Yuan-Hong Liao, Sven Elflein, Liu He, Laura Leal-Taixé, Yejin Choi, Sanja Fidler, David Acuna |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.15362v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2504.15362)| [//]: #04/23\n|[THOUGHTTERMINATOR: Benchmarking, Calibrating, and Mitigating Overthinking in Reasoning Models](https://arxiv.org/abs/2504.13367) \u003cbr\u003e Xiao Pu, Michael Saxon, Wenyue Hua, William Yang Wang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.13367v1/x2.png\"\u003e |[Paper](https://arxiv.org/abs/2504.13367)| [//]: #04/21\n|[Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs](https://arxiv.org/abs/2412.21187) \u003cbr\u003e 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, Dong Yu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2412.21187v2/x2.png\"\u003e |[Paper](https://arxiv.org/abs/2412.21187)|[//]: #03/16\n|[The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks](https://arxiv.org/abs/2502.08235) \u003cbr\u003e Alejandro Cuadron, Dacheng Li, Wenjie Ma, Xingyao Wang, Yichuan Wang, Siyuan Zhuang, Shu Liu, Luis Gaspar Schroeder, Tian Xia, Huanzhi Mao, Nicholas Thumiger, Aditya Desai, Ion Stoica, Ana Klimovic, Graham Neubig, Joseph E. Gonzalez |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/openhands.png\"\u003e |[Paper](https://arxiv.org/abs/2502.08235)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/divelab/sys2bench.svg?style=social\u0026label=Star)](https://github.com/divelab/sys2bench)\u003cbr\u003e[Inference-Time Computations for LLM Reasoning and Planning: A Benchmark and Insights](https://arxiv.org/abs/2502.12521) \u003cbr\u003e Shubham Parashar, Blake Olson, Sambhav Khurana, Eric Li, Hongyi Ling, James Caverlee, Shuiwang Ji |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.12521v1/x1.png\"\u003e |[Github](https://github.com/divelab/sys2bench) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.12521)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/usail-hkust/benchmark_inference_time_computation_LLM.svg?style=social\u0026label=Star)](https://github.com/usail-hkust/benchmark_inference_time_computation_LLM)\u003cbr\u003e[Bag of Tricks for Inference-time Computation of LLM Reasoning](https://arxiv.org/abs/2502.07191) \u003cbr\u003e Fan Liu, Wenshuo Chao, Naiqiang Tan, Hao Liu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.07191v4/x1.png\"\u003e |[Github](https://github.com/usail-hkust/benchmark_inference_time_computation_LLM) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.07191)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/RyanLiu112/compute-optimal-tts.svg?style=social\u0026label=Star)](https://github.com/RyanLiu112/compute-optimal-tts)\u003cbr\u003e[Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling](https://arxiv.org/abs/2502.06703) \u003cbr\u003e Runze Liu, Junqi Gao, Jian Zhao, Kaiyan Zhang, Xiu Li, Biqing Qi, Wanli Ouyang, Bowen Zhou |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.06703v1/x2.png\"\u003e |[Github](https://github.com/RyanLiu112/compute-optimal-tts) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.06703)| [//]: #04/08\n|[DNA Bench: When Silence is Smarter -- Benchmarking Over-Reasoning in Reasoning LLMs](https://arxiv.org/abs/2503.15793) \u003cbr\u003e Masoud Hashemi, Oluwanifemi Bamgbose, Sathwik Tejaswi Madhusudhan, Jishnu Sethumadhavan Nair, Aman Tiwari, Vikas Yadav |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.15793v3/x5.png\"\u003e |[Paper](https://arxiv.org/abs/2503.15793)| [//]: #04/08\n|[S1-Bench: A Simple Benchmark for Evaluating System 1 Thinking Capability of Large Reasoning Models](https://arxiv.org/abs/2504.10368) \u003cbr\u003e Wenyuan Zhang, Shuaiyi Nie, Xinghua Zhang, Zefeng Zhang, Tingwen Liu |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/S1_Bench.png\"\u003e |[Paper](https://arxiv.org/abs/2504.10368)| [//]: #04/14\n|[![Star](https://img.shields.io/github/stars/zhishuifeiqian/VCR-Bench.svg?style=social\u0026label=Star)](https://github.com/zhishuifeiqian/VCR-Bench)\u003cbr\u003e[VCR-Bench: A Comprehensive Evaluation Framework for Video Chain-of-Thought Reasoning](https://arxiv.org/abs/2504.07956) \u003cbr\u003e Yukun Qi, Yiming Zhao, Yu Zeng, Xikun Bao, Wenxuan Huang, Lin Chen, Zehui Chen, Jie Zhao, Zhongang Qi, Feng Zhao |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/video.png\"\u003e |[Github](https://github.com/zhishuifeiqian/VCR-Bench) \u003cbr\u003e [Paper](https://arxiv.org/abs/2504.07956)| [//]: #04/16\n\n\n### Background Papers\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[![Star](https://img.shields.io/github/stars/LeapLabTHU/limit-of-RLVR.svg?style=social\u0026label=Star)](https://github.com/LeapLabTHU/limit-of-RLVR)\u003cbr\u003e[Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?](https://arxiv.org/abs/2504.13837) \u003cbr\u003e Yang Yue, Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang, Yang Yue, Shiji Song, Gao Huang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2504.13837v1/x1.png\"\u003e |[Github](https://github.com/LeapLabTHU/limit-of-RLVR) \u003cbr\u003e [Paper](https://arxiv.org/abs/2504.13837)| [//]: #04/22\n| [![Publish](https://img.shields.io/badge/Conference-NeurIPS_2022-blue)]()\u003cbr\u003e[Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://arxiv.org/abs/2201.11903) \u003cbr\u003e Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, Denny Zhou |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/cot_prompting.png\"\u003e |[Paper](https://arxiv.org/abs/2201.11903)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/princeton-nlp/tree-of-thought-llm.svg?style=social\u0026label=Star)](https://github.com/princeton-nlp/tree-of-thought-llm) [![Publish](https://img.shields.io/badge/Conference-NeurIPS_2023-blue)]()\u003cbr\u003e[Tree of Thoughts: Deliberate Problem Solving with Large Language Models](https://arxiv.org/abs/2305.10601) \u003cbr\u003e Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, Karthik Narasimhan |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2305.10601v2/x1.png\"\u003e |[Github](https://github.com/princeton-nlp/tree-of-thought-llm) \u003cbr\u003e [Paper](https://arxiv.org/abs/2305.10601)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/spcl/graph-of-thoughts.svg?style=social\u0026label=Star)](https://github.com/spcl/graph-of-thoughts) [![Publish](https://img.shields.io/badge/Conference-AAAI_2024-blue)]()\u003cbr\u003e[Graph of Thoughts: Solving Elaborate Problems with Large Language Models](https://arxiv.org/abs/2308.09687) \u003cbr\u003e Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, Torsten Hoefler |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/got.png\"\u003e |[Github](https://github.com/spcl/graph-of-thoughts) \u003cbr\u003e [Paper](https://arxiv.org/abs/2308.09687)| [//]: #04/08\n| [![Publish](https://img.shields.io/badge/Conference-ICLR_2023-blue)]()\u003cbr\u003e[Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://arxiv.org/abs/2203.11171) \u003cbr\u003e Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, Denny Zhou |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/sc.png\"\u003e |[Paper](https://arxiv.org/abs/2203.11171)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/TIGER-AI-Lab/Program-of-Thoughts.svg?style=social\u0026label=Star)](https://github.com/TIGER-AI-Lab/Program-of-Thoughts) [![Publish](https://img.shields.io/badge/Conference-TMLR_2023-blue)]()\u003cbr\u003e[Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks](https://arxiv.org/abs/2211.12588) \u003cbr\u003e Wenhu Chen, Xueguang Ma, Xinyi Wang, William W. Cohen |\u003cimg width=\"1002\" alt=\"image\" src=\"figures/pot.png\"\u003e |[Github](https://github.com/TIGER-AI-Lab/Program-of-Thoughts) \u003cbr\u003e [Paper](https://arxiv.org/abs/2211.12588)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/hanxuhu/chain-of-symbol-planning.svg?style=social\u0026label=Star)](https://github.com/hanxuhu/chain-of-symbol-planning) [![Publish](https://img.shields.io/badge/Conference-COLM_2024-blue)]()\u003cbr\u003e[Chain-of-Symbol Prompting Elicits Planning in Large Langauge Models](https://arxiv.org/abs/2305.10276) \u003cbr\u003e Hanxu Hu, Hongyuan Lu, Huajian Zhang, Yun-Ze Song, Wai Lam, Yue Zhang |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2305.10276v7/x1.png\"\u003e |[Github](https://github.com/hanxuhu/chain-of-symbol-planning) \u003cbr\u003e [Paper](https://arxiv.org/abs/2305.10276)| [//]: #04/08\n\n###### Survey\n| Title \u0026 Authors | Introduction | Links |\n|:--|  :----: | :---:|\n|[Thinking Machines: A Survey of LLM based Reasoning Strategies](https://arxiv.org/abs/2503.10814) \u003cbr\u003e Dibyanayan Bandyopadhyay, Soham Bhattacharjee, Asif Ekbal |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2503.10814v1/x1.png\"\u003e |[Paper](https://arxiv.org/abs/2503.10814)| [//]: #04/08\n|[![Star](https://img.shields.io/github/stars/zzli2022/Awesome-System2-Reasoning-LLM.svg?style=social\u0026label=Star)](https://github.com/zzli2022/Awesome-System2-Reasoning-LLM)\u003cbr\u003e[From System 1 to System 2: A Survey of Reasoning Large Language Models](https://arxiv.org/abs/2502.17419) \u003cbr\u003e 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, Cheng-Lin Liu |\u003cimg width=\"1002\" alt=\"image\" src=\"https://arxiv.org/html/2502.17419v2/extracted/6232702/images/timeline.png\"\u003e |[Github](https://github.com/zzli2022/Awesome-System2-Reasoning-LLM) \u003cbr\u003e [Paper](https://arxiv.org/abs/2502.17419)| [//]: #04/08\n\n\n\n## Acknowledgement\n\nThis repository is inspired by [Awesome-Efficient-LLM](https://github.com/horseee/Awesome-Efficient-LLM/)\n\n\n## Citation\n\n```bibtex\n@article{,\n    title={Efficient Reasoning Models: A Survey},\n    author={Feng, Sicheng and Fang, Gongfan and Ma, Xinyin and Wang, Xinchao},\n    journal={arXiv preprint arXiv:2504.10903},\n    year={2025},\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ffscdc%2FAwesome-Efficient-Reasoning-Models","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ffscdc%2FAwesome-Efficient-Reasoning-Models","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ffscdc%2FAwesome-Efficient-Reasoning-Models/lists"}