Elamur Research⌕
☰
home / papers / 2024 / Artifacts or Abduction: How Do LLMs Answer Multiple-Choice Questions Without the Question?

Artifacts or Abduction: How Do LLMs Answer Multiple-Choice Questions Without the Question?

Paper · 2024 · Nishant Balepur, Abhilasha Ravichander and Rachel Rudinger

Summary

Multiple-choice question answering (MCQA) is widely used to rank LLMs, but it is unclear whether high scores reflect the skills these benchmarks are meant to test. The authors check whether models can succeed when the question is hidden. They give LLaMA-2 70B, Falcon 40B, Mixtral 8x7B and Phi-2 "choices-only" few-shot prompts on ARC-Challenge, MMLU and HellaSwag. In 11 of 12 model–dataset pairs, these prompts score clearly above a majority baseline.

To explain this, they run a black-box analysis of three possible sources:

  • Memorization: they find no sign that recalled test items account for the gains.
  • Per-choice priors: judging each option in isolation does not fully explain the gains, which suggests models compare options against each other.
  • Abductive question inference: models can reconstruct a plausible question from the options, sometimes the original one, but this also does not fully explain the results.

The study matters for benchmark design. It argues for reporting choices-only baselines, building more artifact-resistant MCQA datasets, and probing LLM decision-making further.

Key points

  • The largest choices-only gain over the majority baseline was about 0.33 accuracy, for LLaMA-2 on HellaSwag. This required few exemplars: 5 per subject for MMLU, 10 for HellaSwag and 25 for ARC.
  • For memorization, prompts with uninformative choices never beat the majority baseline by a large margin. Invalid outputs were scored as random guesses (0.25).
  • The two-step abductive question inference (AQI) prompt performed about as well as the choices-only prompt. It outperformed choices-only for 3 LLMs on HellaSwag, and the two prompts showed moderate agreement.
  • For LLaMA-2 on ARC, there were 43 cases where AQI chose the gold answer and its inferred question was answerable by one of the choices. In 42% of these, the inferred question matched the meaning of the original.
  • The evaluation suite and data are released at github.com/nbalepur/mcqa-artifacts.
  • Closed-source models were not tested, and the analysis did not inspect logits.

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

🔬Allen Institute for Artificial IntelligenceSeattle · United States275 works · 97,634 citations🎓University of Maryland, College ParkCollege Park · United States132 works · 25,380 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 worksComputer Vision and Pattern Recognition · 2,869 works

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

Cite

Cite as Nishant Balepur, Abhilasha Ravichander and Rachel Rudinger. 2024. Artifacts or Abduction: How Do LLMs Answer Multiple-Choice Questions Without the Question?. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2024.acl-long.555

@inproceedings{balepur2024artifacts,
  title = {{Artifacts or Abduction: How Do LLMs Answer Multiple-Choice Questions Without the Question?}},
  author = {Nishant Balepur, Abhilasha Ravichander and Rachel Rudinger},
  booktitle = {Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2024},
  doi = {10.18653/v1/2024.acl-long.555},
  url = {https://doi.org/10.18653/v1/2024.acl-long.555}
}

References

63 references, 30 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Marah Abdin, Jyoti Aneja, Sebastien Bubeck, Caio César Teodoro Mendes, Weizhu Chen, Allie Del Giorno, Ronen Eldan, Sivakanth Gopi, Suriya Gunasekar, Mojan Javaheripi, Piero Kauffmann, Yin Tat Lee, Yuanzhi Li, Anh Nguyen, Gustavo de Rosa, Olli Saarikivi, Adil Salim, Shital Shah, Michael Santacroce, H… find
  2. Norah Alzahrani, Hisham Abdullah Alyahya, Yazeed Alnumay, Sultan Alrashed, Shaykhah Alsubaie, Yusef Almushaykeh, Faisal Mirza, Nouf Alotaibi, Nora Altwairesh, Areeb Alowisheq, et al. 2024. When benchmarks are targets: Revealing the sensitivity of large language model leaderboards. arXiv preprint arX… find
  3. Yonatan Belinkov, Adam Poliak, Stuart Shieber, Benjamin Van Durme, and Alexander Rush. 2019. Don’t take the premise for granted: Mitigating artifacts in natural language inference. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 877–891, Florence, It…
  4. Anat Ben-Simon, David V Budescu, and Baruch Nevo. 1997. A comparative study of measures of partial knowledge in multiple-choice tests. Applied Psychological Measurement, 21(1):65–88. find
  5. Chandra Bhagavatula, Ronan Le Bras, Chaitanya Malaviya, Keisuke Sakaguchi, Ari Holtzman, Hannah Rashkin, Doug Downey, Wen tau Yih, and Yejin Choi. 2020. Abductive commonsense reasoning. In International Conference on Learning Representations.
  6. Nicholas Carlini, Florian Tramèr, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom B. Brown, Dawn Xiaodong Song, Úlfar Erlingsson, Alina Oprea, and Colin Raffel. 2020. Extracting training data from large language models. In USENIX Security Symposium.
  7. Dhawaleswar Rao Ch and Sujan Kumar Saha. 2018. Automatic multiple choice question generation from text: A survey. IEEE Transactions on Learning Technologies, 13(1):14–25. find
  8. Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. 2018. Think you have solved question answering? try arc, the ai2 reasoning challenge. arXiv preprint arXiv:1803.05457.
  9. Jacob Cohen. 1960. A coefficient of agreement for nominal scales. Educational and psychological measurement, 20(1):37–46. find
  10. Mengnan Du, Fengxiang He, Na Zou, Dacheng Tao, and Xia Hu. 2023. Shortcut learning of large language models in natural language understanding. Communications of the ACM, 67(1):110–120. find
  11. Yanai Elazar, Bhargavi Paranjape, Hao Peng, Sarah Wiegreffe, Khyathi Raghavi, Vivek Srikumar, Sameer Singh, and Noah A Smith. 2023. Measuring and improving attentiveness to partial inputs with counterfactuals. arXiv preprint arXiv:2311.09605. find
  12. Nishant Balepur, Shramay Palta, and Rachel Rudinger. 2023. It’s not easy being wrong: Evaluating process of elimination reasoning in large language models. arXiv preprint arXiv:2311.07532. — [arXiv:2311.07532](https://arxiv.org/abs/2311.07532) find
  13. Shi Feng, Eric Wallace, and Jordan Boyd-Graber. 2019. Misleading failures of partial-input baselines. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5533– 5538, Florence, Italy. Association for Computational Linguistics. find
  14. Edward Beeching, Clémentine Fourrier, Nathan Habib, Sheon Han, Nathan Lambert, Nazneen Rajani, Omar Sanseviero, Lewis Tunstall, and Thomas Wolf. 2023. Open llm leaderboard. https://huggingface.co/spaces/ HuggingFaceH4/open_llm_leaderboard. ↗
  15. Matt Gardner, Yoav Artzi, Victoria Basmov, Jonathan Berant, Ben Bogin, Sihao Chen, Pradeep Dasigi, Dheeru Dua, Yanai Elazar, Ananth Gottumukkala, Nitish Gupta, Hannaneh Hajishirzi, Gabriel Ilharco, Daniel Khashabi, Kevin Lin, Jiangming Liu, Nelson F. Liu, Phoebe Mulcaire, Qiang Ning, Sameer Singh, N… — Evaluating Models' Local Decision Boundaries via Contrast Sets
  16. Mor Geva, Yoav Goldberg, and Jonathan Berant. 2019. Are we modeling the task or the annotator? an investigation of annotator bias in natural language understanding datasets. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conf… find
  17. Max Glockner, Vered Shwartz, and Yoav Goldberg. 2018. Breaking NLI systems with sentences that require simple lexical inferences. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 650–655, Melbourne, Australia. Association for…
  18. Yash Goyal, Tejas Khot, Douglas Summers-Stay, Dhruv Batra, and Devi Parikh. 2017. Making the v in vqa matter: Elevating the role of image understanding in visual question answering. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 6904–6913.
  19. Suchin Gururangan, Swabha Swayamdipta, Omer Levy, Roy Schwartz, Samuel Bowman, and Noah A. Smith. 2018. Annotation artifacts in natural language inference data. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Techno…
  20. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring massive multitask language understanding. In International Conference on Learning Representations.
  21. Christine Herlihy and Rachel Rudinger. 2021. MedNLI is not immune: Natural language inference artifacts in the clinical domain. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volum…
  22. Alexander Hoyle, Rupak Sarkar, Pranav Goel, and Philip Resnik. 2023. Natural language decompositions of implicit content enable better text representations. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 13188–13214, Singapore. Association for Comput… find
  23. Jie Huang and Kevin Chen-Chuan Chang. 2023. Towards reasoning in large language models: A survey. In Findings of the Association for Computational Linguistics: ACL 2023, pages 1049–1065, Toronto, Canada. Association for Computational Linguistics.
  24. Jie Huang, Hanyin Shao, and Kevin Chen-Chuan Chang. 2022. Are large pre-trained language models leaking your personal information? In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 2038–2047, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. find
  25. Shotaro Ishihara. 2023. Training data extraction from pre-trained language models: A survey. In Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP 2023), pages 260– 275, Toronto, Canada. Association for Computational Linguistics. find
  26. Robin Jia and Percy Liang. 2017. Adversarial examples for evaluating reading comprehension systems. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2021–2031, Copenhagen, Denmark. Association for Computational Linguistics.
  27. Albert Q. Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, Blanche Savary, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Emma Bou Hanna, Florian Bressand, Gianna Lengyel, Guillaume Bour, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock,… find
  28. Jaehun Jung, Lianhui Qin, Sean Welleck, Faeze Brahman, Chandra Bhagavatula, Ronan Le Bras, and Yejin Choi. 2022. Maieutic prompting: Logically consistent reasoning with recursive explanations. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 1266–1279,…
  29. Divyansh Kaushik and Zachary C. Lipton. 2018. How much reading does reading comprehension require? a critical investigation of popular benchmarks. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 5010– 5015, Brussels, Belgium. Association for Computati…
  30. Takeshi Kojima, Shixiang (Shane) Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2022. Large language models are zero-shot reasoners. In Advances in Neural Information Processing Systems, volume 35, pages 22199–22213. Curran Associates, Inc.
  31. Hector J. Levesque, Ernest Davis, and L. Morgenstern. 2011. The winograd schema challenge. In AAAI Spring Symposium: Logical Formalizations of Commonsense Reasoning. find
  32. Percy Liang, Rishi Bommasani, Tony Lee, Dimitris Tsipras, Dilara Soylu, Michihiro Yasunaga, Yian Zhang, Deepak Narayanan, Yuhuai Wu, Ananya Kumar, Benjamin Newman, Binhang Yuan, Bobby Yan, Ce Zhang, Christian Alexander Cosgrove, Christopher D Manning, Christopher Re, Diana Acosta-Navas, Drew Arad Hu… find
  33. Alisa Liu, Swabha Swayamdipta, Noah A. Smith, and Yejin Choi. 2022. WANLI: Worker and AI collaboration for natural language inference dataset creation. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 6826–6847, Abu Dhabi, United Arab Emirates. Association for Computat…
  34. Chenkai Ma and Xinya Du. 2023. POE: Process of elimination for multiple choice reasoning. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 4487–4496, Singapore. Association for Computational Linguistics. find
  35. R. Thomas McCoy, Ellie Pavlick, and Tal Linzen. 2019. Right for the wrong reasons: Diagnosing syntactic heuristics in natural language inference. In Annual Meeting of the Association for Computational Linguistics.
  36. Michail Mersinias and Panagiotis Valvis. 2022. Mitigating dataset artifacts in natural language inference through automatic contextual data augmentation and learning optimization. In International Conference on Language Resources and Evaluation.
  37. Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, and Luke Zettlemoyer. 2022. Rethinking the role of demonstrations: What makes in-context learning work? In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 11048–11064, Ab…
  38. John Morris, Eli Lifland, Jin Yong Yoo, Jake Grigsby, Di Jin, and Yanjun Qi. 2020. TextAttack: A framework for adversarial attacks, data augmentation, and adversarial training in NLP. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pa…
  39. Jane Pan, Tianyu Gao, Howard Chen, and Danqi Chen. 2023. What in-context learning "learns" in-context: Disentangling task recognition and task learning. In Annual Meeting of the Association for Computational Linguistics.
  40. Mihir Parmar, Swaroop Mishra, Mor Geva, and Chitta Baral. 2023. Don’t blame the annotator: Bias already starts in the annotation instructions. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 1779–1789, Dubrovnik, Croatia. Associat… find
  41. Charles Sanders Peirce. 1974. Collected papers of charles sanders peirce, volume 1. Harvard University Press. find
  42. Guilherme Penedo, Quentin Malartic, Daniel Hesslow, Ruxandra Cojocaru, Hamza Alobeidli, Alessandro Cappelli, Baptiste Pannier, Ebtesam Almazrouei, and Julien Launay. 2023. The refinedweb dataset for falcon LLM: Outperforming curated corpora with web data only. In Thirty-seventh Conference on Neural… find
  43. Pouya Pezeshkpour and Estevam Hruschka. 2023. Large language models sensitivity to the order of options in multiple-choice questions. ArXiv, abs/2308.11483.
  44. Adam Poliak, Jason Naradowsky, Aparajita Haldar, Rachel Rudinger, and Benjamin Van Durme. 2018. Hypothesis only baselines in natural language inference. In Proceedings of the Seventh Joint Conference on Lexical and Computational Semantics, pages 180–191, New Orleans, Louisiana. Association for Compu…
  45. Colin Raffel, Noam M. Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2019. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21:140:1–140:67.
  46. Abhilasha Ravichander, Joe Stacey, and Marek Rei. 2023. When and why does bias mitigation work? In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9233–9247, Singapore. Association for Computational Linguistics. find
  47. Joshua Robinson and David Wingate. 2023. Leveraging large language models for multiple choice question answering. In The Eleventh International Conference on Learning Representations.
  48. Kenneth D Royal and Mari-Wells Hedgpeth. 2015. A novel method for evaluating examination item quality. International Journal of Psychological Studies, 7(1):17. find
  49. Kenneth D Royal and Myrah R Stockdale. 2017. The impact of 3-option responses to multiple-choice questions on guessing strategies and cut score determinations. Journal of Advances in Medical Education & Professionalism, 5(2):84. find
  50. Oscar Sainz, Jon Ander Campos, Iker García-Ferrero, Julen Etxaniz, and Eneko Agirre. 2023. Did chatgpt cheat on your test? find
  51. Krunal Shah, Nitish Gupta, and Dan Roth. 2020. What do we expect from multiple-choice QA systems? In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 3547–3553, Online. Association for Computational Linguistics. find
  52. Neha Srikanth and Rachel Rudinger. 2022. Partial-input baselines show that NLI models can ignore context, but they don’t. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 4753–4763, Seattle, Unite…
  53. Rowan Zellers, Ari Holtzman, Yonatan Bisk, Ali Farhadi, and Yejin Choi. 2019. Hellaswag: Can a machine really finish your sentence? In Annual Meeting of the Association for Computational Linguistics. find
  54. Chujie Zheng, Hao Zhou, Fandong Meng, Jie Zhou, and Minlie Huang. 2024. Large language models are not robust multiple choice selectors. In The Twelfth International Conference on Learning Representations. find
  55. Kun Zhou, Yutao Zhu, Zhipeng Chen, Wentong Chen, Wayne Xin Zhao, Xu Chen, Yankai Lin, Jinhui Wen, and Jiawei Han. 2023. Don’t make your llm an evaluation benchmark cheater. ArXiv, abs/2311.01964. find
  56. Saku Sugawara, Pontus Stenetorp, Kentaro Inui, and Akiko Aizawa. 2019. Assessing the benchmarking capacity of machine reading comprehension datasets. In AAAI Conference on Artificial Intelligence. find
  57. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288.
  58. Miles Turpin, Julian Michael, Ethan Perez, and Sam Bowman. 2023. Language models don’t always say what they think: Unfaithful explanations in chain-of-thought prompting. ArXiv, abs/2305.04388.
  59. Eric Wallace, Shi Feng, Nikhil Kandpal, Matt Gardner, and Sameer Singh. 2019. Universal adversarial triggers for attacking and analyzing nlp. In Conference on Empirical Methods in Natural Language Processing. find
  60. Haochun Wang, Sendong Zhao, Zewen Qiang, Bing Qin, and Ting Liu. 2024. Beyond the answers: Reviewing the rationality of multiple choice question answering for the evaluation of large language models. arXiv preprint arXiv:2402.01349. — [arXiv:2402.01349](https://arxiv.org/abs/2402.01349) find
  61. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35:24824–24837.
  62. Sarah Wiegreffe and Ana Marasović. 2021. Teach me to explain: A review of datasets for explainable natural language processing. In NeurIPS Datasets and Benchmarks. find
  63. Yue Yu, Yuchen Zhuang, Jieyu Zhang, Yu Meng, Alexander Ratner, Ranjay Krishna, Jiaming Shen, and Chao Zhang. 2023. Large language model as attributed training data generator: A tale of diversity and bias. In Thirty-Seventh Conference on Neural Information Processing Systems Datasets and Benchmarks T…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/balepur-2024-artifacts-or-abduction-how-do/