Elamur Research⌕
☰
home / papers / 2022 / Faithful Reasoning Using Large Language Models

Faithful Reasoning Using Large Language Models

Paper · 2022 · Antonia Creswell and Murray Shanahan

Summary

Large language models usually answer questions in a single opaque pass. Users cannot see which premises were used or check the logic, and multi-step problems suffer. Creswell and Shanahan (DeepMind) propose a forward-chaining system whose computation follows the formal definition of a valid reasoning trace.

Its core, Selection-Inference (SI), alternates two fine-tuned LMs:

  • Selection picks supporting statements from the given context by their sentence labels, so it cannot invent facts.
  • Inference derives a new statement from the selection alone, without seeing the question.

A Halter LM decides after each step whether the question can be answered and maps the latest inference onto the required answer format. A learned value function scores partial traces to steer a step-level beam search.

On Proof Writer and a QA version of EntailmentBank, the system beats baselines on final-answer accuracy. It also produces more accurate, checkable traces, hallucinates less and relies more on the context. This matters for trust and debugging, because users can audit each step instead of accepting an unexplained answer.

Key points

  • The architecture has four separately fine-tuned LMs: Selection, Inference, Halter and value function. None is trained end-to-end for final-answer accuracy.
  • Connectedness of the trace is guaranteed because selections must come from context labels such as "sent 3". Validity holds only if each individual inference step is logically correct.
  • If no answer is reached within a fixed step budget, the model returns "Unknown". Filtering out these abstentions noticeably raises precision.
  • Beam search expands P candidate steps per trace, keeps the top B ≤ P by value score, and continues until all traces have halted. Search improved both SI and the Proof Only baseline.
  • Limitations: the context is assumed to be supplied, and step correctness is not formally guaranteed. Future work aims to add retrieval to populate the context, framed as an instance of "algorithmic prompting".

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 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

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

Cite

Cite as Antonia Creswell and Murray Shanahan. 2022. Faithful Reasoning Using Large Language Models. arXiv:2208.14271.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2208.14271 · arXiv 2208.14271

@misc{creswell2022faithful,
  title = {{Faithful Reasoning Using Large Language Models}},
  author = {Antonia Creswell and Murray Shanahan},
  year = {2022},
  doi = {10.48550/arxiv.2208.14271},
  eprint = {2208.14271},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2208.14271}
}

References

31 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. J.-B. Alayrac, J. Donahue, P. Luc, A. Miech, I. Barr, Y. Hasson, K. Lenc, A. Mensch, K. Millican, M. Reynolds, et al. Flamingo: a visual language model for few-shot learning. arXiv preprint arXiv:2204.14198, 2022.
  2. A. Banino, J. Balaguer, and C. Blundell. Pondernet: Learning to ponder. arXiv preprint arXiv:2107.05407, 2021. find
  3. E. M. Bender, T. Gebru, A. McMillan-Major, and S. Shmitchell. On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, FAccT ’21, page 610–623, New York, NY, USA, 2021. Association for Computing Mac… — [doi:10.1145/3442188.3445922](https://doi.org/10.1145/3442188.3445922) find
  4. G. Betz, C. Voigt, and K. Richardson. Critical thinking for language models. In Proceedings of the 14th International Conference on Computational Semantics (IWCS), pages 63– 75, Groningen, The Netherlands (online), June 2021. Association for Computational Linguistics. URL https://aclanthology.org/ 2… ↗
  5. K. Bostrom, Z. Sprague, S. Chaudhuri, and G. Durrett. Natural language deduction through search over statement compositions. arXiv preprint arXiv:2201.06028, 2022. find
  6. P. Clark, I. Cowhey, O. Etzioni, T. Khot, A. Sabharwal, C. Schoenick, and O. Tafjord. Think you have solved question answering? try arc, the ai2 reasoning challenge. arXiv preprint arXiv:1803.05457, 2018.
  7. K. Cobbe, V. Kosaraju, M. Bavarian, J. Hilton, R. Nakano, C. Hesse, and J. Schulman. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168, 2021.
  8. A. Creswell, M. Shanahan, and I. Higgins. Selection-inference: Exploiting large language models for interpretable logical reasoning. arXiv preprint arXiv:2205.09712, 2022. find
  9. B. Dalvi, P. Jansen, O. Tafjord, Z. Xie, H. Smith, L. Pipatanangkura, and P. Clark. Explaining answers with entailment trees. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 7358–7370, Online and Punta Cana, Dominican Republic, Nov. 2021. Association…
  10. B. Dalvi, O. Tafjord, and P. Clark. Towards teachable reasoning systems. arXiv preprint arXiv:2204.13074, 2022. find
  11. I. Dasgupta, A. K. Lampinen, S. C. Chan, A. Creswell, D. Kumaran, J. L. McClelland, and F. Hill. Language models show human-like content effects on reasoning. arXiv preprint arXiv:2207.07051, 2022.
  12. A. Graves. Generating sequences with recurrent neural networks. arXiv preprint arXiv:1308.0850, 2013. find
  13. A. Graves. Adaptive computation time for recurrent neural networks. arXiv preprint arXiv:1603.08983, 2016. find
  14. T. Kojima, S. S. Gu, M. Reid, Y. Matsuo, and Y. Iwasawa. Large language models are zero-shot reasoners. arXiv preprint arXiv:2205.11916, 2022.
  15. S. Kumar and P. Talukdar. Nile: Natural language inference with faithful natural language explanations. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8730–8742, 2020.
  16. R. Nakano, J. Hilton, S. Balaji, J. Wu, L. Ouyang, C. Kim, C. Hesse, S. Jain, V. Kosaraju, W. Saunders, et al. Webgpt: Browser-assisted question-answering with human feedback. arXiv preprint arXiv:2112.09332, 2021.
  17. V. Gupta, S. Zhang, A. Vempala, Y. He, T. Choji, and V. Srikumar. Right for the right reason: Evidence extraction for trustworthy tabular reasoning. In Proceedings of the 60th Annual M. Nye, A. J. Andreassen, G. Gur-Ari, H. Michalewski, J. Austin, D. Bieber, D. Dohan, Meeting of the Association for… find
  18. Casas, L. A. Hendricks, J. Welbl, A. Clark, et al. Training compute-optimal large language models. arXiv preprint arXiv:2203.15556, 2022.
  19. H. Jhamtani and P. Clark. Learning to explain: Datasets and models for identifying valid reasoning chains in multihop question-answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 137–150, Online, Nov. 2020. Association for Computational… find
  20. S. Kadavath, T. Conerly, A. Askell, T. Henighan, D. Drain, E. Perez, N. Schiefer, Z. H. Dodds, N. DasSarma, E. Tran-Johnson, et al. Language models (mostly) know what they know. arXiv preprint arXiv:2207.05221, 2022. find
  21. D. Ribeiro, S. Wang, X. Ma, R. Dong, X. Wei, H. Zhu, X. Chen, Z. Huang, P. Xu, A. Arnold, et al. Entailment tree explanations via iterative retrieval-generation reasoner. arXiv preprint arXiv:2205.09224, 2022. find
  22. S. Saha, S. Ghosh, S. Srivastava, and M. Bansal. Prover: Proof generation for interpretable reasoning over rules. In EMNLP (1), 2020.
  23. T. Sellam, D. Das, and A. Parikh. Bleurt: Learning robust metrics for text generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7881–7892, 2020.
  24. O. Tafjord, B. Dalvi, and P. Clark. Proofwriter: Generating implications, proofs, and abductive statements over natural language. In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 3621–3634, 2021.
  25. J. Wei, X. Wang, D. Schuurmans, M. Bosma, E. Chi, Q. Le, and D. Zhou. Chain of thought prompting elicits reasoning in large language models, 2022. — Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  26. Uesato, Jordi Grau-Moya, Ramana Kumar and Irina Higgins for insightful discussions. L. Weidinger, J. Mellor, M. Rauh, C. Griffin, J. Uesato, P. Huang, M. Cheng, M. Glaese, B. Balle, A. Kasirzadeh, Z. Kenton, S. Brown, W. Hawkins, T. Stepleton, C. Biles, A. Birhane, J. Haas, L. Rimell, L. A. Hendrick… — Ethical and social risks of harm from Language Models
  27. Z. Xie, S. Thiem, J. Martin, E. Wainwright, S. Marmorstein, and P. Jansen. Worldtree v2: A corpus of science-domain structured explanations and inference patterns supporting multi-hop inference. In Proceedings of the 12th Language Resources and Evaluation Conference, pages 5456– 5473, 2020. find
  28. E. Zelikman, Y. Wu, and N. D. Goodman. Star: Bootstrapping reasoning with reasoning. arXiv preprint arXiv:2203.14465, 2022.
  29. A. Zeng, A. Wong, S. Welker, K. Choromanski, F. Tombari, A. Purohit, M. Ryoo, V. Sindhwani, J. Lee, V. Vanhoucke, et al. Socratic models: Composing zero-shot multimodal reasoning with language. arXiv preprint arXiv:2204.00598, 2022.
  30. H. Zhang, L. H. Li, T. Meng, K.-W. Chang, and G. V. d. Broeck. On the paradox of learning to reason from data. arXiv preprint arXiv:2205.11502, 2022. — [arXiv:2205.11502](https://arxiv.org/abs/2205.11502) find
  31. D. Zhou, N. Schärli, L. Hou, J. Wei, N. Scales, X. Wang, D. Schuurmans, O. Bousquet, Q. Le, and E. Chi. Least-to-most prompting enables complex reasoning in large language models. arXiv preprint arXiv:2205.10625, 2022. \#\# Acknowledgements The authors would like to thank Angeliki Lazaridou, Charles B…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/creswell-2022-faithful-reasoning-using-large-language/