Elamur Research⌕
☰
home / papers / 2023 / Large Language Models are In-Context Semantic Reasoners rather than Symbolic Reasoners

Large Language Models are In-Context Semantic Reasoners rather than Symbolic Reasoners

Paper · 2023 · Xiaojuan Tang, Zilong Zheng, Jiaqi Li et al.

Summary

This work asks where the apparent few-shot reasoning skill of LLMs actually comes from. The authors argue that it depends largely on token meanings absorbed during pretraining, which let models link concepts into shallow shortcut chains instead of applying rules formally. To test this, they remove or invert semantics in reasoning problems by swapping natural entities and relations for abstract symbols or for counter-commonsense content. They then probe three abilities: deduction (judging whether a derived fact is true), induction (writing a rule from example facts and a template) and abduction (picking facts plus a rule that explain an observation). Memorization of newly supplied knowledge is also examined. For evaluation they introduce Symbolic Tree, a synthetic, closed-world, noise-free set of multi-hop problems generated from logical rules, and they also use ProofWriter. Models do reasonably well when the content matches everyday knowledge. They falter when they must rely only on unfamiliar knowledge given in the prompt. This challenges claims of human-like formal reasoning in LLMs. It also motivates semantics-free benchmarks and pairing LLMs with external knowledge stores.

Key points

  • Evaluation uses Symbolic Tree and ProofWriter. Metrics are accuracy for deduction, exact-match precision against gold rules for induction, and Proof Accuracy (PA) for abduction. Code is released as ICSR on GitHub.
  • Zero-shot-CoT beats plain zero-shot on symbolic versions of the tasks but does worse in the semantics-rich setting.
  • In the "Removing rules" setting, the model sees only the relevant facts and must rely on its internal knowledge. This setting scored higher than supplying explicit logical rules.
  • On commonsense benchmarks, GPT-4 zero-shot reaches 82.1/79.4 on CSQA dev/test and 91.8 on OpenBookQA. ChatGPT zero-shot gets 71.9/67.8/70.8, and GreaseLM gets 78.5/74.2/84.8. CoT-style few-shot prompting lowers ChatGPT's CSQA dev score to 58.9 (predict-explain) or 46.4 (explain-predict).
  • The memorization ability of LLMs falls short of graph-based methods. The authors suggest combining LLMs with non-parametric knowledge bases such as graph databases.

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

🎓Peking UniversityBeijing · China244 works · 39,359 citations🏛️Beijing Institute for General Artificial IntelligenceBeijing · China7 works · 576 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

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

Cite

Cite as Xiaojuan Tang, Zilong Zheng, Jiaqi Li, Fanxu Meng, Song-Chun Zhu, Yitao Liang and Muhan Zhang. 2023. Large Language Models are In-Context Semantic Reasoners rather than Symbolic Reasoners. arXiv:2305.14825.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2305.14825 · arXiv 2305.14825

@misc{tang2023large,
  title = {{Large Language Models are In-Context Semantic Reasoners rather than Symbolic Reasoners}},
  author = {Xiaojuan Tang, Zilong Zheng, Jiaqi Li, Fanxu Meng, Song-Chun Zhu, Yitao Liang and Muhan Zhang},
  year = {2023},
  doi = {10.48550/arxiv.2305.14825},
  eprint = {2305.14825},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2305.14825}
}

References

63 references, 31 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020.
  2. Zhuosheng Zhang, Aston Zhang, Mu Li, and Alex Smola. Automatic chain of thought prompting in large language models. arXiv preprint arXiv:2210.03493, 2022.
  3. Rishi Bommasani, Drew A Hudson, Ehsan Adeli, Russ Altman, Simran Arora, Sydney von Arx, Michael S Bernstein, Jeannette Bohg, Antoine Bosselut, Emma Brunskill, et al. On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258, 2021.
  4. Kevin Lu, Aditya Grover, Pieter Abbeel, and Igor Mordatch. Pretrained transformers as universal computation engines. arXiv preprint arXiv:2103.05247, 1, 2021. — [arXiv:2103.05247](https://arxiv.org/abs/2103.05247) find
  5. 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.
  6. 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.
  7. Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, and Luke Zettlemoyer. Rethinking the role of demonstrations: What makes in-context learning work? arXiv preprint arXiv:2202.12837, 2022.
  8. Jiachang Liu, Dinghan Shen, Yizhe Zhang, Bill Dolan, Lawrence Carin, and Weizhu Chen. What makes good in-context examples for gpt-3? arXiv preprint arXiv:2101.06804, 2021.
  9. Zihao Zhao, Eric Wallace, Shi Feng, Dan Klein, and Sameer Singh. Calibrate before use: Improving few-shot performance of language models. In International Conference on Machine Learning, pages 12697–12706. PMLR, 2021.
  10. Timo Schick and Hinrich Schütze. Exploiting cloze questions for few shot text classification and natural language inference. arXiv preprint arXiv:2001.07676, 2020.
  11. Yao Lu, Max Bartolo, Alastair Moore, Sebastian Riedel, and Pontus Stenetorp. Fantastically ordered prompts and where to find them: Overcoming few-shot prompt order sensitivity. arXiv preprint arXiv:2104.08786, 2021.
  12. Ekin Akyürek, Dale Schuurmans, Jacob Andreas, Tengyu Ma, and Denny Zhou. What learning algorithm is in-context learning? investigations with linear models. arXiv preprint arXiv:2211.15661, 2022.
  13. Sang Michael Xie, Aditi Raghunathan, Percy Liang, and Tengyu Ma. An explanation of in-context learning as implicit bayesian inference. arXiv preprint arXiv:2111.02080, 2021.
  14. Steven A Sloman. The empirical case for two systems of reasoning. Psychological bulletin, 119(1):3, 1996. find
  15. Thomas Gilovich, Dale Griffin, and Daniel Kahneman. Heuristics and biases: The psychology of intuitive judgment. Cambridge university press, 2002. find
  16. Oyvind Tafjord, Bhavana Dalvi Mishra, and Peter Clark. Proofwriter: Generating implications, proofs, and abductive statements over natural language. arXiv preprint arXiv:2012.13048, 2020.
  17. Robin Cooper. The fracas consortium robin cooper, dick crouch, jan van eijck, chris fox, josef van genabith, jan jaspars, hans kamp, david milward, manfred pinkal, massimo poesio, steve pulman. 1996. find
  18. SamuelR. Bowman, Gabor Angeli, Christopher Potts, and ChristopherD. Manning. A large annotated corpus for learning natural language inference, Aug 2015.
  19. Adina Williams, Nikita Nangia, and SamuelR. Bowman. A broad-coverage challenge corpus for sentence understanding through inference, Apr 2017.
  20. Jean-Philippe Bernardy and Stergios Chatzikyriakidis. Improving the precision of natural textual entailment problem datasets., May 2020.
  21. Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. Commonsenseqa: A question answering challenge targeting commonsense knowledge, Nov 2018.
  22. Shourya Aggarwal, Divyanshu Mandowara, Vishwajeet Agrawal, Dinesh Khandelwal, Parag Singla, and Dinesh Garg. Explanations for commonsenseqa: New dataset and models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference…
  23. Dheeru Dua, Yizhong Wang, Pradeep Dasigi, Gabriel Stanovsky, Sameer Singh, and Matt Gardner. Drop: A reading comprehension benchmark requiring discrete reasoning over paragraphs, Mar 2019.
  24. Wenhu Chen, Hanwen Zha, Zhiyu Chen, Wenhan Xiong, Hong Wang, and WilliamYang Wang. Hybridqa: A dataset of multi-hop question answering over tabular and textual data, Apr 2020.
  25. 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.
  26. Subhro Roy and Dan Roth. Solving general arithmetic word problems. arXiv preprint arXiv:1608.01413, 2016.
  27. Mor Geva, Daniel Khashabi, Elad Segal, Tushar Khot, Dan Roth, and Jonathan Berant. Did aristotle use a laptop? a question answering benchmark with implicit reasoning strategies. Transactions of the Association for Computational Linguistics, 9:346–361, 2021. find
  28. Rajarshi Das, Manzil Zaheer, Dung Thai, Ameya Godbole, Ethan Perez, Jay Yoon Lee, Lizhen Tan, Lazaros Polymenakos, and Andrew McCallum. Case-based reasoning for natural language queries over knowledge bases. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, p… find
  29. Ohad Rubin, Jonathan Herzig, and Jonathan Berant. Learning to retrieve prompts for in-context learning. arXiv preprint arXiv:2112.08633, 2021.
  30. Hongjin Su, Jungo Kasai, Chen Henry Wu, Weijia Shi, Tianlu Wang, Jiayi Xin, Rui Zhang, Mari Ostendorf, Luke Zettlemoyer, Noah A. Smith, and Tao Yu. Selective annotation makes language models better few-shot learners, 2022. find
  31. Victor Sanh, Albert Webson, Colin Raffel, Stephen H. Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Teven Le Scao, Arun Raja, Manan Dey, M Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Nihal Nayak, Debajyoti Da… — Multitask Prompted Training Enables Zero-Shot Task Generalization
  32. Jason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, and Quoc V. Le. Finetuned language models are zero-shot learners, 2022.
  33. Yao Lu, Max Bartolo, AlastairP. Moore, Sebastian Riedel, and Pontus Stenetorp. Fantastically ordered prompts and where to find them: Overcoming few-shot prompt order sensitivity, Apr 2021.
  34. Albert Webson and Ellie Pavlick. Do prompt-based models really understand the meaning of their prompts? In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 2300–2344, Seattle, United States, July 202…
  35. Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, and Luke Zettlemoyer. Rethinking the role of demonstrations: What makes in-context learning work?
  36. Junyeob Kim, Hyuhng Joon Kim, Hyunsoo Cho, Hwiyeol Jo, Sang-Woo Lee, Sang-goo Lee, Kang Min Yoo, and Taeuk Kim. Ground-truth labels matter: A deeper look into input-label demonstrations. arXiv preprint arXiv:2205.12685, 2022.
  37. Ekin Akyürek, Dale Schuurmans, Jacob Andreas, Tengyu Ma, and Denny Zhou. What learning algorithm is in-context learning? investigations with linear models, Nov 2022.
  38. John McCarthy and Patrick J Hayes. Some philosophical problems from the standpoint of artificial intelligence. In Readings in artificial intelligence, pages 431–450. Elsevier, 1981. find
  39. Nada Lavrac and Saso Dzeroski. Inductive logic programming. In WLP, pages 146–160. Springer, 1994. find
  40. Allen Newell and Herbert A Simon. Computer science as empirical inquiry: Symbols and search. In ACM Turing award lectures, page 1975. 2007. find
  41. Philip N Johnson-Laird. Deductive reasoning. Annual review of psychology, 50(1):109–135, 1999. find
  42. Gyöngyi Kovács and Karen M Spens. Abductive reasoning in logistics research. International journal of physical distribution & logistics management, 2005. find
  43. George Boole. The mathematical analysis of logic. Philosophical Library, 1847. find
  44. John McCarthy. Recursive functions of symbolic expressions and their computation by machine, part i. Communications of the ACM, 3(4):184–195, 1960. find
  45. Norbert Fuhr. Probabilistic datalog: Implementing logical information retrieval for advanced applications. Journal of the American Society for Information Science, 51(2):95–110, 2000. find
  46. Thomas Eiter, Giovambattista Ianni, and Thomas Krennwallner. Answer set programming: A primer. Springer, 2009. find
  47. Kexin Yi, Jiajun Wu, Chuang Gan, Antonio Torralba, Pushmeet Kohli, and Josh Tenenbaum. Neural-symbolic vqa: Disentangling reasoning from vision and language understanding. Advances in neural information processing systems, 31, 2018. find
  48. Eui Chul Shin, Illia Polosukhin, and Dawn Song. Improving neural program synthesis with inferred execution traces. Advances in Neural Information Processing Systems, 31, 2018. find
  49. Guillaume Lample and François Charton. Deep learning for symbolic mathematics. arXiv preprint arXiv:1912.01412, 2019. find
  50. Vishal Pallagani, Bharath Muppasani, Keerthiram Murugesan, Francesca Rossi, Lior Horesh, Biplav Srivastava, Francesco Fabiano, and Andrea Loreggia. Plansformer: Generating symbolic plans using transformers. arXiv preprint arXiv:2212.08681, 2022. — [arXiv:2212.08681](https://arxiv.org/abs/2212.08681) find
  51. Peter Cathcart Wason and Philip Nicholas Johnson-Laird. Psychology of reasoning: Structure and content, volume 86. Harvard University Press, 1972. find
  52. P. C. Wason. Reasoning about a rule. Quarterly Journal of Experimental Psychology, page 273–281, Jul 2007. doi: 10.1080/14640746808400161. URL http://dx.doi.org/10.1080/14640746808400161. — [doi:10.1080/14640746808400161](https://doi.org/10.1080/14640746808400161) ↗
  53. Ronald Fagin, Joseph Y Halpern, Yoram Moses, and Moshe Vardi. Reasoning about knowledge. MIT press, 2004. find
  54. Douglas N Walton. What is reasoning? what is an argument? The journal of Philosophy, 87(8):399–419, 1990. find
  55. Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. Translating embeddings for modeling multi-relational data. Advances in neural information processing systems, 26, 2013.
  56. Patrick Hohenecker and Thomas Lukasiewicz. Ontology reasoning with deep neural networks. Journal of Artificial Intelligence Research, 68:503–540, 2020. find
  57. Massimo Carro. Nosql databases. arXiv preprint arXiv:1401.2101, 2014. — [arXiv:1401.2101](https://arxiv.org/abs/1401.2101) find
  58. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. Stanford alpaca: An instruction-following llama model. https://github.com/tatsu-lab/stanford_alpaca, 2023. ↗
  59. Luis Galárraga, Christina Teflioudi, Katja Hose, and Fabian M Suchanek. Fast rule mining in ontological knowledge bases with amie + +. The VLDB Journal, 24(6):707–730, 2015. find
  60. Jiuhai Chen, Lichang Chen, Heng Huang, and Tianyi Zhou. When do you need chain-of-thought prompting for chatgpt? arXiv preprint arXiv:2304.03262, 2023. — [arXiv:2304.03262](https://arxiv.org/abs/2304.03262) find
  61. Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9, 2019. find
  62. Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. Can a suit of armor conduct electricity? a new dataset for open book question answering. arXiv preprint arXiv:1809.02789, 2018. find
  63. Xikun Zhang, Antoine Bosselut, Michihiro Yasunaga, Hongyu Ren, Percy Liang, Christopher D Manning, and Jure Leskovec. Greaselm: Graph reasoning enhanced language models for question answering. arXiv preprint arXiv:2201.08860, 2022. Table of Contents A Prompts find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tang-2023-large-language-models-context-semantic/