Elamur Research⌕
☰
home / papers / 2023 / MathChat: Converse to Tackle Challenging Math Problems with LLM Agents

MathChat: Converse to Tackle Challenging Math Problems with LLM Agents

Paper · 2023 · Yiran Wu; Feiran Jia; Shaokun Zhang; Hangyu Li; Erkang Zhu; Yue Wang; Yin Tat Lee; Richard Peng; Qingyun Wu; and Chi Wang

Summary

The paper asks whether hard competition mathematics can be handled better when an LLM solves a problem through a conversation with another agent rather than in a single response. The authors present MathChat, a framework that pairs an LLM agent (GPT-4) with a user proxy agent. The proxy sends the problem together with an initial prompt, runs any Python code it finds in the LLM's replies, and returns outputs or error messages. It also nudges the model with replies such as "Continue", or asks it to solve the problem itself when errors keep recurring. Code that ran successfully earlier is re-executed alongside each new snippet, so the step-by-step state is preserved. The dialogue ends once a final answer appears inside \boxed{}. Evaluation uses the level-5 (hardest) problems of the MATH dataset. MathChat is compared with two code-based prompting baselines, Program of Thoughts (PoT) and Program Synthesis, plus a vanilla prompt. MathChat outperforms the earlier tool-using prompts by about 6%. The work suggests that multi-turn agent dialogue with code execution helps multi-step reasoning.

Key points

  • Two agents: an LLM agent (GPT-4) and a user proxy agent that executes tools (Python) and provides guidance across multi-turn dialogue.
  • Benchmark: level-5 problems from the MATH dataset (Hendrycks et al., 2021), which are difficult high school competition problems.
  • Baselines: PoT (Chen et al., 2022), Program Synthesis (Drori et al., 2022), and a vanilla prompt.
  • MathChat improves on prior tool-using prompting methods by about 6%. It reaches 60% accuracy in half of the problem categories and stays competitive in the rest.
  • The authors show that MathChat can be extended with other prompts and tools, analyse failure causes for all methods, and include a case where GPT-4 fails because it chose direct reasoning over code.

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

🏢MicrosoftRedmond · United States325 works · 142,931 citations🎓University of WashingtonSeattle · United States353 works · 200,043 citations🎓University of WaterlooWaterloo · Canada67 works · 13,592 citations🎓Imperial College LondonLondon · United Kingdom48 works · 12,181 citations🎓Pennsylvania State UniversityState College · United States56 works · 5,360 citations

 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 worksComputational Theory and Mathematics · 252 works

Cite

Cite as Yiran Wu; Feiran Jia; Shaokun Zhang; Hangyu Li; Erkang Zhu; Yue Wang; Yin Tat Lee; Richard Peng; Qingyun Wu; and Chi Wang. 2023. MathChat: Converse to Tackle Challenging Math Problems with LLM Agents. arXiv:2306.01337.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2306.01337 · arXiv 2306.01337

@misc{wu2023mathchat,
  title = {{MathChat: Converse to Tackle Challenging Math Problems with LLM Agents}},
  author = {Yiran Wu; Feiran Jia; Shaokun Zhang; Hangyu Li; Erkang Zhu; Yue Wang; Yin Tat Lee; Richard Peng; Qingyun Wu; and Chi Wang},
  year = {2023},
  doi = {10.48550/arxiv.2306.01337},
  eprint = {2306.01337},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2306.01337}
}

References

51 references, 24 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. BabyAGI. Github — babyagi. https://github.com/yoheinakajima/babyagi, 2023. ↗
  2. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Michal Podstawski, Hubert Niewiadomski, Piotr Nyczyk, et al. Graph of thoughts: Solving elaborate problems with large language models. arXiv preprint arXiv:2308.09687, 2023. find
  3. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, et al. Sparks of artificial general intelligence: Early experiments with gpt-4. arXiv preprint arXiv:2303.12712, 2023.
  4. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374, 2021.
  5. Wenhu Chen, Xueguang Ma, Xinyi Wang, and William W Cohen. Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks. arXiv preprint arXiv:2211.12588, 2022. find
  6. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, et al. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168, 2021.
  7. Antonia Creswell, Murray Shanahan, and Irina Higgins. Selection-inference: Exploiting large language models for interpretable logical reasoning. arXiv preprint arXiv:2205.09712, 2022. find
  8. Iddo Drori, Sarah Zhang, Reece Shuttleworth, Leonard Tang, Albert Lu, Elizabeth Ke, Kevin Liu, Linda Chen, Sunny Tran, Newman Cheng, et al. A neural network solves, explains, and generates university math problems by program synthesis and few-shot learning at human level. Proceedings of the National… find
  9. Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. Improving factuality and reasoning in language models through multiagent debate. arXiv preprint arXiv:2305.14325, 2023.
  10. Dheeru Dua, Shivanshu Gupta, Sameer Singh, and Matt Gardner. Successive prompting for decomposing complex questions. arXiv preprint arXiv:2212.04092, 2022. find
  11. Luyu Gao, Aman Madaan, Shuyan Zhou, Uri Alon, Pengfei Liu, Yiming Yang, Jamie Callan, and Graham Neubig. Pal: Program-aided language models. arXiv preprint arXiv:2211.10435, 2022. find
  12. 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. arXiv preprint arXiv:2103.03874, 2021.
  13. Sirui Hong, Xiawu Zheng, Jonathan Chen, Yuheng Cheng, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, et al. Metagpt: Meta programming for multi-agent collaborative framework. arXiv preprint arXiv:2308.00352, 2023.
  14. Tushar Khot, Harsh Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Ashish Sabharwal. Decomposed prompting: A modular approach for solving complex tasks. arXiv preprint arXiv:2210.02406, 2022. find
  15. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. Large language models are zero-shot reasoners. arXiv preprint arXiv:2205.11916, 2022.
  16. Guohao Li, Hasan Abed Al Kader Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. Camel: Communicative agents for ”mind” exploration of large scale language model society, 2023. find
  17. Yifei Li, Zeqi Lin, Shizhuo Zhang, Qiang Fu, Bei Chen, Jian-Guang Lou, and Weizhu Chen. On the advance of making language models better reasoners. arXiv preprint arXiv:2206.02336, 2022. — [arXiv:2206.02336](https://arxiv.org/abs/2206.02336) find
  18. Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Zhaopeng Tu, and Shuming Shi. Encouraging divergent thinking in large language models through multi-agent debate, 2023.
  19. Jieyi Long. Large language model guided tree-of-thought. arXiv preprint arXiv:2305.08291, 2023. find
  20. Marvin Minsky. Society of mind. Simon and Schuster, 1988. find
  21. Ansong Ni, Srini Iyer, Dragomir Radev, Ves Stoyanov, Wen-tau Yih, Sida I Wang, and Xi Victoria Lin. Lever: Learning to verify language-to-code generation with execution. arXiv preprint arXiv:2302.08468, 2023.
  22. OpenAI. Gpt-4 technical report, 2023.
  23. Bhargavi Paranjape, Scott Lundberg, Sameer Singh, Hannaneh Hajishirzi, Luke Zettlemoyer, and Marco Tulio Ribeiro. Art: Automatic multi-step reasoning and tool-use for large language models. arXiv preprint arXiv:2303.09014, 2023. find
  24. Aaron Parisi, Yao Zhao, and Noah Fiedel. Talm: Tool augmented language models. arXiv preprint arXiv:2205.12255, 2022.
  25. George Polya. How to solve it: A new aspect of mathematical method. Number 246. Princeton university press, 2004. find
  26. Ofir Press, Muru Zhang, Sewon Min, Ludwig Schmidt, Noah A Smith, and Mike Lewis. Measuring and narrowing the compositionality gap in language models. arXiv preprint arXiv:2210.03350, 2022. find
  27. Timo Schick, Jane Dwivedi-Yu, Roberto Dessı̀, Roberta Raileanu, Maria Lomeli, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. Toolformer: Language models can teach themselves to use tools. arXiv preprint arXiv:2302.04761, 2023.
  28. Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. arXiv preprint arXiv:1701.06538, 2017. find
  29. Yongliang Shen, Kaitao Song, Xu Tan, Dongsheng Li, Weiming Lu, and Yueting Zhuang. Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face. Advances in Neural Information Processing Systems, 36, 2024.
  30. Linxin Song, Jiale Liu, Jieyu Zhang, Shaokun Zhang, Ao Luo, Shijian Wang, Qingyun Wu, and Chi Wang. Adaptive in-conversation team building for language model agents. arXiv preprint arXiv:2405.19425, 2024. — [arXiv:2405.19425](https://arxiv.org/abs/2405.19425) find
  31. Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, et al. A survey on large language model based autonomous agents. arXiv preprint arXiv:2308.11432, 2023a.
  32. Lei Wang, Wanyu Xu, Yihuai Lan, Zhiqiang Hu, Yunshi Lan, Roy Ka-Wei Lee, and Ee-Peng Lim. Plan-and-solve prompting: Improving zero-shot chain-of-thought reasoning by large language models. arXiv preprint arXiv:2305.04091, 2023b. find
  33. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171, 2022.
  34. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. Chain of thought prompting elicits reasoning in large language models. arXiv preprint arXiv:2201.11903, 2022.
  35. Yixuan Weng, Minjun Zhu, Shizhu He, Kang Liu, and Jun Zhao. Large language models are reasoners with self-verification. arXiv preprint arXiv:2212.09561, 2022.
  36. Eugene P Wigner. The unreasonable effectiveness of mathematics in the natural sciences. In Mathematics and science, pp. 291–306. World Scientific, 1990. find
  37. Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Shaokun Zhang, Erkang Zhu, Beibin Li, Li Jiang, Xiaoyun Zhang, and Chi Wang. Autogen: Enabling next-gen llm applications via multi-agent conversation framework. arXiv preprint arXiv:2308.08155, 2023.
  38. Tongshuang Wu, Ellen Jiang, Aaron Donsbach, Jeff Gray, Alejandra Molina, Michael Terry, and Carrie J Cai. Promptchainer: Chaining large language model prompts through visual programming. In CHI Conference on Human Factors in Computing Systems Extended Abstracts, pp. 1–10, 2022. find
  39. Yiran Wu, Tianwei Yue, Shaokun Zhang, Chi Wang, and Qingyun Wu. Stateflow: Enhancing llm task-solving through state-driven workflows. arXiv preprint arXiv:2403.11322, 2024.
  40. Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, et al. The rise and potential of large language model based agents: A survey. arXiv preprint arXiv:2309.07864, 2023.
  41. Jingfeng Yang, Haoming Jiang, Qingyu Yin, Danqing Zhang, Bing Yin, and Diyi Yang. Seqzero: Few-shot compositional semantic parsing with sequential prompts and zero-shot models. arXiv preprint arXiv:2205.07381, 2022.
  42. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629, 2022.
  43. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate problem solving with large language models. arXiv preprint arXiv:2305.10601, 2023. find
  44. Shaokun Zhang, Feiran Jia, Chi Wang, and Qingyun Wu. Targeted hyperparameter optimization with lexicographic preferences over multiple objectives. In The Eleventh international conference on learning representations, 2023a. find
  45. Shaokun Zhang, Yiran Wu, Zhonghua Zheng, Qingyun Wu, and Chi Wang. Hypertime: Hyperparameter optimization for combating temporal distribution shifts. arXiv preprint arXiv:2305.18421, 2023b. — [arXiv:2305.18421](https://arxiv.org/abs/2305.18421) find
  46. Shaokun Zhang, Xiaobo Xia, Zhaoqing Wang, Ling-Hao Chen, Jiale Liu, Qingyun Wu, and Tongliang Liu. Ideal: Influence-driven selective annotations empower in-context learners in large language models. arXiv preprint arXiv:2310.10873, 2023c. — [arXiv:2310.10873](https://arxiv.org/abs/2310.10873) find
  47. Shaokun Zhang, Jieyu Zhang, Jiale Liu, Linxin Song, Chi Wang, Ranjay Krishna, and Qingyun Wu. Training language model agents without modifying language models. arXiv preprint arXiv:2402.11359, 2024. find
  48. Yifan Zhang, Jingqin Yang, Yang Yuan, and Andrew Chi-Chih Yao. Cumulative reasoning with large language models. arXiv preprint arXiv:2308.04371, 2023d. find
  49. Chuanyang Zheng, Zhengying Liu, Enze Xie, Zhenguo Li, and Yu Li. Progressive-hint prompting improves reasoning in large language models. arXiv preprint arXiv:2304.09797, 2023. find
  50. Aojun Zhou, Ke Wang, Zimu Lu, Weikang Shi, Sichun Luo, Zipeng Qin, Shaoqing Lu, Anya Jia, Linqi Song, Mingjie Zhan, et al. Solving challenging math word problems using gpt-4 code interpreter with code-based self-verification. arXiv preprint arXiv:2308.07921, 2023.
  51. Denny Zhou, Nathanael Schärli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Olivier Bousquet, Quoc Le, and Ed Chi. Least-to-most prompting enables complex reasoning in large language models. arXiv preprint arXiv:2205.10625, 2022.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wu-2023-mathchat-converse-tackle-challenging-math/