Elamur Research⌕
☰
home / papers / 2025 / Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching

Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching

Paper · 2025 · Simon A. Aytes, Jinheon Baek and Sung Ju Hwang

Summary

Chain-of-Thought (CoT) prompting helps large language models solve multi-step problems. However, its long natural-language traces increase output length, latency and cost. The authors propose Sketch-of-Thought (SoT), a training-free prompting framework that has models write compressed, structured reasoning instead of full explanatory sentences. SoT draws on cognitive science and offers three paradigms:

  • Conceptual Chaining: short links between key concepts, motivated by associative memory.
  • Chunked Symbolism: dense, equation-like steps, motivated by working-memory chunking.
  • Expert Lexicons: domain abbreviations and notation, motivated by research on expert schemas.

A lightweight router model examines each query and selects the paradigm at inference time. SoT was evaluated on 18 reasoning benchmarks covering mathematics, commonsense, logic, multi-hop, science and medicine, plus multilingual and multimodal settings. It reduces output tokens by up to 84% relative to CoT with little accuracy loss, and it sometimes improves accuracy in mathematical and multi-hop tasks. This makes reasoning LLMs more practical under tight compute or budget limits.

Key points

  • Paradigm–task matching: Chunked Symbolism targets math and arithmetic. Conceptual Chaining targets commonsense, multi-hop, logical and scientific questions. Expert Lexicons targets specialised domains such as medicine (e.g., STEMI/MONA-style shorthand).
  • Baselines: SoT is compared with CoT, Constrained CoT (CCoT) and Chain-of-Draft (CoD). Models shown in the main comparison include GPT-4o, Sonnet 3.5, Qwen 7B/14B/32B and Llama 8B/11B.
  • Design choices: The router was trained on annotations produced by GPT-4o, so it may inherit that model's biases. Each paradigm uses a fixed set of exemplars.
  • Compatibility: SoT works alongside ensemble prompting strategies. Code is released at github.com/SimonAytes/SoT.
  • Limitations: Multilingual tests covered only widely spoken languages, so effects on low-resource languages remain untested. Code generation and other new paradigms are left for future work.

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

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,320 worksReasoning · 1,426 works

Cite

Cite as Simon A. Aytes, Jinheon Baek and Sung Ju Hwang. 2025. Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.emnlp-main.1236

@inproceedings{aytes2025sketch,
  title = {{Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching}},
  author = {Simon A. Aytes, Jinheon Baek and Sung Ju Hwang},
  booktitle = {Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing},
  year = {2025},
  doi = {10.18653/v1/2025.emnlp-main.1236},
  url = {https://doi.org/10.18653/v1/2025.emnlp-main.1236}
}

References

53 references, 27 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. John R. Anderson. 1983. A spreading activation theory of memory. Journal of Verbal Learning and Verbal Behavior, 22(3):261–295. find
  2. Anthropic. 2024. Claude 3.5 sonnet. find
  3. Daman Arora and Andrea Zanette. 2025. Training language models to reason efficiently. Preprint, arXiv:2502.04463. find
  4. A. Baddeley. 2000. The episodic buffer: a new component of working memory? Trends in Cognitive Sciences, 4(11):417–423. find
  5. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, and Torsten Hoefler. 2024. Graph of thoughts: Solving elaborate problems with large language models. Proceedings of the AAAI Conference… find
  6. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, Harsha Nori, Hamid Palangi, Marco Tulio Ribeiro, and Yi Zhang. 2023. Sparks of artificial general intelligence: Early experiments with gpt-4. Preprint, a…
  7. Michelene T. H. Chi, Paul J. Feltovich, and Robert Glaser. 1981. Categorization and Representation of Physics Problems by Experts and Novices. Cognitive Science, 5(2):121–152.
  8. 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. 2021. Training Verifiers to Solve Math Word Problems. arXiv preprint. ArXiv:2110.14168 [cs].
  9. Tri Dao. 2023. Flashattention-2: Faster attention with better parallelism and work partitioning. Preprint, arXiv:2307.08691.
  10. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. BERT: pre-training of deep bidirectional transformers for language understanding. CoRR, abs/1810.04805.
  11. Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. 2023. Improving factuality and reasoning in language models through multiagent debate. Preprint, arXiv:2305.14325.
  12. Dheeru Dua, Yizhong Wang, Pradeep Dasigi, Gabriel Stanovsky, Sameer Singh, and Matt Gardner. 2019. DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs. arXiv preprint. ArXiv:1903.00161 [cs].
  13. Mor Geva, Daniel Khashabi, Elad Segal, Tushar Khot, Dan Roth, and Jonathan Berant. 2021. Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategies. arXiv preprint. ArXiv:2101.02235 [cs]. find
  14. Vinod Goel. 1995. Sketches of Thought. MIT Press. find
  15. Shibo Hao, Sainbayar Sukhbaatar, DiJia Su, Xian Li, Zhiting Hu, Jason Weston, and Yuandong Tian. 2024. Training large language models to reason in a continuous latent space. Preprint, arXiv:2412.06769. find
  16. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring Massive Multitask Language Understanding. arXiv preprint. ArXiv:2009.03300 [cs].
  17. Xijie Huang, Li Lyna Zhang, Kwang-Ting Cheng, Fan Yang, and Mao Yang. 2024. Fewer is more: Boosting llm reasoning with reinforced context pruning. Preprint, arXiv:2312.08901. — [arXiv:2312.08901](https://arxiv.org/abs/2312.08901) find
  18. Drew A Hudson and Christopher D Manning. 2019. Gqa: A new dataset for real-world visual reasoning and compositional question answering. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6700–6709.
  19. Peter A. Jansen, Elizabeth Wainwright, Steven Marmorstein, and Clayton T. Morrison. 2018. WorldTree: A Corpus of Explanation Graphs for Elementary Science Questions supporting Multi-Hop Inference. arXiv preprint. ArXiv:1802.03052 [cs].
  20. Huiqiang Jiang, Qianhui Wu, Chin-Yew Lin, Yuqing Yang, and Lili Qiu. 2023. Llmlingua: Compressing prompts for accelerated inference of large language models. Preprint, arXiv:2310.05736.
  21. Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. 2020. What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams. arXiv preprint. ArXiv:2009.13081 [cs].
  22. Qiao Jin, Bhuwan Dhingra, Zhengping Liu, William Cohen, and Xinghua Lu. 2019. PubMedQA: A Dataset for Biomedical Research Question Answering. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Proce…
  23. Jaap Jumelet, Willem Zuidema, and Arabella Sinclair. 2024. Do language models exhibit human-like structural priming effects? Preprint, arXiv:2406.04847. find
  24. Tushar Khot, Peter Clark, Michal Guerquin, Peter Jansen, and Ashish Sabharwal. 2020. QASC: A Dataset for Question Answering via Sentence Composition. arXiv preprint. ArXiv:1910.11473 [cs]. find
  25. Wang Ling, Dani Yogatama, Chris Dyer, and Phil Blunsom. 2017. Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problems. arXiv preprint. ArXiv:1705.04146 [cs]. find
  26. Jian Liu, Leyang Cui, Hanmeng Liu, Dandan Huang, Yile Wang, and Yue Zhang. 2020. LogiQA: A Challenge Dataset for Machine Reading Comprehension with Logical Reasoning. arXiv preprint. ArXiv:2007.08124 [cs].
  27. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-eval: NLG evaluation using gpt-4 with better human alignment. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 2511–2522, Singapore. Association for Computational Lin…
  28. Pan Lu, Swaroop Mishra, Tony Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Oyvind Tafjord, Peter Clark, and Ashwin Kalyan. 2022. Learn to explain: Multimodal reasoning via thought chains for science question answering. In The 36th Conference on Neural Information Processing Systems (NeurIPS). find
  29. George A. Miller. 1956. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2):81–97. Place: US Publisher: American Psychological Association. find
  30. Sania Nayab, Giulio Rossolini, Marco Simoni, Andrea Saracino, Giorgio Buttazzo, Nicolamaria Manes, and Fabrizio Giacomelli. 2025. Concise thoughts: Impact of output length on llm reasoning and cost. Preprint, arXiv:2407.19825. — [arXiv:2407.19825](https://arxiv.org/abs/2407.19825) find
  31. OpenAI. 2024. Gpt-4o system card. arXiv:2410.21276.
  32. Preprint, Arkil Patel, Satwik Bhattamishra, and Navin Goyal. 2021. Are NLP Models really able to Solve Simple Math Word Problems? In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 2080–2094, Online…
  33. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. find
  34. Leonardo Ranaldi and Andre Freitas. 2024. Self-refine instruction-tuning for aligning reasoning in language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, page 2325–2347. Association for Computational Linguistics.
  35. Matthew Renze and Erhan Guven. 2024. The benefits of a concise chain of thought on problem-solving in large language models. Preprint, arXiv:2401.05618. — [arXiv:2401.05618](https://arxiv.org/abs/2401.05618) find
  36. Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2020. Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter. Preprint, arXiv:1910.01108. find
  37. Zayne Sprague, Fangcong Yin, Juan Diego Rodriguez, Dongwei Jiang, Manya Wadhwa, Prasann Singhal, Xinyu Zhao, Xi Ye, Kyle Mahowald, and Greg Durrett. 2024. To cot or not to cot? chain-of-thought helps mainly on math and symbolic reasoning. Preprint, arXiv:2409.12183.
  38. Meta. 2024b. Llama 3.2: Revolutionizing edge ai and vision with open, customizable models. find
  39. Jiankai Sun, Chuanyang Zheng, Enze Xie, Zhengying Liu, Ruihang Chu, Jianing Qiu, Jiaqi Xu, Mingyu Ding, Hongyang Li, Mengzhe Geng, Yue Wu, Wenhai Wang, Junsong Chen, Zhangyue Yin, Xiaozhe Ren, Jie Fu, Junxian He, Wu Yuan, Qi Liu, Xihui Liu, Yu Li, Hao Dong, Yu Cheng, Ming Zhang, Pheng Ann Heng, Jife… — [arXiv:2312.11562](https://arxiv.org/abs/2312.11562) find
  40. Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. 2018. Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering. arXiv preprint. ArXiv:1809.02789 [cs]. find
  41. Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. 2019. CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge. In Proceedings of the 2019 Conference Meta. 2024a. Introducing llama 3.1: Our most capable models to date. of the North American Chapter of the Ass…
  42. Qwen Team. 2024. Qwen2.5: A party of foundation models. find
  43. Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2022. MuSiQue: Multi-hop Questions via Single-hop Question Composition. Transactions of the Association for Computational Linguistics, 10:539–554. Place: Cambridge, MA Publisher: MIT Press. find
  44. Weihao Yu, Zihang Jiang, Yanfei Dong, and Jiashi Feng. 2020. ReClor: A Reading Comprehension Dataset Requiring Logical Reasoning. arXiv preprint. ArXiv:2002.04326 [cs].
  45. Murong Yue, Jie Zhao, Min Zhang, Liang Du, and Ziyu Yao. 2024. Large language model cascades with mixture of thoughts representations for cost-efficient reasoning. Preprint, arXiv:2310.03094. — [arXiv:2310.03094](https://arxiv.org/abs/2310.03094) find
  46. Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, Yifan Du, Chen Yang, Yushuo Chen, Zhipeng Chen, Jinhao Jiang, Ruiyang Ren, Yifan Li, Xinyu Tang, Zikang Liu, Peiyu Liu, Jian-Yun Nie, and Ji-Rong Wen. 2024. A survey of l…
  47. Peifeng Wang, Zhengyang Wang, Zheng Li, Yifan Gao, Bing Yin, and Xiang Ren. 2023a. Scott: Self-consistent chain-of-thought distillation. Preprint, arXiv:2305.01879.
  48. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2023b. Self-consistency improves chain of thought reasoning in language models. Preprint, arXiv:2203.11171.
  49. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. 2023. Chain-of-thought prompting elicits reasoning in large language models. Preprint, arXiv:2201.11903.
  50. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier-ric 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
  51. Silei Xu, Wenhao Xie, Lingxiao Zhao, and Pengcheng He. 2025. Chain of draft: Thinking faster by writing less. Preprint, arXiv:2502.18600. find
  52. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W. Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering. arXiv preprint. ArXiv:1809.09600 [cs].
  53. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. Preprint, arXiv:2305.10601. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/aytes-2025-sketch-thought-efficient-llm-reasoning/