Elamur Research⌕
☰
home / papers / 2023 / DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering

DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering

Paper · 2023 · Ella Neeman, Roee Aharoni, Or Honovich et al.

Summary

Generative question answering systems rely on two kinds of information: facts memorized in their weights during pretraining, and evidence supplied in an input passage. When the two are mixed, a user cannot tell which one produced an answer. This hurts trust and makes it hard to handle cases where outdated memory contradicts the given text.

The authors train a single model to emit two answers in one generation. The contextual answer is grounded in the passage, and the parametric answer comes from what the model has memorized. Training data is derived from Natural Questions with gold passages and augmented in two ways:

  • Counterfactual contexts: the answer entity in the passage is replaced with a different entity, following Longpre et al. (2021).
  • Answerability examples: the passage is empty or randomly chosen, so the contextual answer must be "unanswerable".

Experiments with T5 models of several sizes show three effects. Disentangled models are more robust to knowledge conflicts, abstain more reliably when the context lacks an answer, and still give useful memory-based answers. This supports attribution, explanation and debugging of QA outputs.

Key points

  • Training uses four example types: factual, counterfactual, empty and random. In all of them the parametric target stays the original NQ answer, and only the context and contextual target change.
  • The data is restricted to NQ items that come with a short answer as well as a gold passage (about 35% of the data), and oracle retrieval is assumed. The training split has 85,540 factual, empty and random examples each, plus 30,653 counterfactual ones. The test split has 1,365 examples per type.
  • The two augmentation strategies proved complementary, and both are needed for proper disentanglement.
  • For T5-11B, validation checkpoints showed a trade-off between factual and random contexts (e.g., 1.02M vs 1.04M steps), driven by how often the model said "no answer". T5-large showed the two tasks moving together more closely.
  • Code and data are released at github.com/ellaneeman/disent_qa.

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

🏢GoogleMountain View · United States353 works · 364,806 citations🎓Tel Aviv UniversityTel Aviv · Israel67 works · 10,006 citations🎓Hebrew University of JerusalemJerusalem · Israel50 works · 9,147 citations

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 9,085 worksArtificial Intelligence · 8,888 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 worksComputer Vision and Pattern Recognition · 1,282 works

🧭 Topics 1Research subjects and languages the work is about.Question answering · 243 works

Cite

Cite as Ella Neeman, Roee Aharoni, Or Honovich, Leshem Choshen, Idan Szpektor and Omri Abend. 2023. DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering. Proceedings of the 61st 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/v1/2023.acl-long.559

@inproceedings{neeman2023disentqa,
  title = {{DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering}},
  author = {Ella Neeman, Roee Aharoni, Or Honovich, Leshem Choshen, Idan Szpektor and Omri Abend},
  booktitle = {Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2023},
  doi = {10.18653/v1/2023.acl-long.559},
  url = {https://doi.org/10.18653/v1/2023.acl-long.559}
}

References

29 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ekin Akyürek, Tolga Bolukbasi, Frederick Liu, Binbin Xiong, Ian Tenney, Jacob Andreas, and Kelvin Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 3932–3945, Online. Association for Computational Linguistics. find
  2. Guu. 2022. Tracing knowledge in language models back to the training data. arXiv preprint arXiv:2205.11482. — [arXiv:2205.11482](https://arxiv.org/abs/2205.11482) find
  3. Akari Asai and Eunsol Choi. 2021. Challenges in information-seeking QA: Unanswerable questions and paragraph retrieval. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Lon… find
  4. Hung-Ting Chen, Michael Zhang, and Eunsol Choi. 2022. Rich knowledge sources bring complex knowledge conflicts: Recalibrating models to reflect conflicting evidence. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 2292–2307, Abu Dhabi, United Arab Emi…
  5. Christopher Clark, Kenton Lee, Ming-Wei Chang, Tom Kwiatkowski, Michael Collins, and Kristina Toutanova. 2019. BoolQ: Exploring the surprising difficulty of natural yes/no questions. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics:…
  6. Damai Dai, Li Dong, Yaru Hao, Zhifang Sui, Baobao Chang, and Furu Wei. 2022. Knowledge neurons in pretrained transformers. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8493–8502, Dublin, Ireland. Association for Computation… find
  7. Nicola De Cao, Wilker Aziz, and Ivan Titov. 2021. Editing factual knowledge in language models. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 6491–6506, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics. find
  8. Bhuwan Dhingra, Jeremy R. Cole, Julian Martin Eisenschlos, Daniel Gillick, Jacob Eisenstein, and William W. Cohen. 2022. Time-aware language models as temporal knowledge bases. Transactions of the Association for Computational Linguistics, 10:257–273.
  9. Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, Kristina Toutanova, Llion Jones, Matthew Kelcey, Ming-Wei Chang, Andrew M. Dai, Jakob Uszkoreit, Quoc Le, and Slav Petrov. 2019. Natural…
  10. Patrick Lewis, Pontus Stenetorp, and Sebastian Riedel. 2021. Question and answer test-train overlap in open-domain question answering datasets. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 1000–1008, Online. Associ… find
  11. Daliang Li, Ankit Singh Rawat, Manzil Zaheer, Xin Wang, Michal Lukasik, Andreas Veit, Felix Yu, and Surinder Kumar. 2022. Large language models with controllable working memory. ArXiv, abs/2211.05110.
  12. Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. TruthfulQA: Measuring how models mimic human falsehoods. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3214–3252, Dublin, Ireland. Association for Computational Linguistics…
  13. Shayne Longpre, Kartik Perisetla, Anthony Chen, Nikhil Ramesh, Chris DuBois, and Sameer Singh. 2021. Entity-based knowledge conflicts in question answering. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 7052–7063, Online and Punta Cana, Dominican Re…
  14. Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov. 2022. Locating and editing factual associations in gpt. In Neural Information Processing Systems. find
  15. Evan Hernandez, Belinda Z Li, and Jacob Andreas. 2023. Measuring and manipulating knowledge representations in language models. arXiv preprint arXiv:2304.00740. find
  16. Fabio Petroni, Aleksandra Piktus, Angela Fan, Patrick Lewis, Majid Yazdani, Nicola De Cao, James Thorne, Yacine Jernite, Vladimir Karpukhin, Jean Maillard, Vassilis Plachouras, Tim Rocktäschel, and Sebastian Riedel. 2021. KILT: a benchmark for knowledge intensive language tasks. In Proceedings of th…
  17. Najoung Kim, Ellie Pavlick, Burcu Karagol Ayan, and Deepak Ramachandran. 2021. Which linguist invented the lightbulb? presupposition verification for question-answering. In Proceedings of the 59th Annual Meeting of the Association for Computational Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, P…
  18. Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2463–2473, Hong Kong, China. Association for Computational Linguistics. find
  19. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learning Research, 21(140):1–67.
  20. Pranav Rajpurkar, Robin Jia, and Percy Liang. 2018. Know what you don’t know: Unanswerable questions for SQuAD. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 784– 789, Melbourne, Australia. Association for Computational Lin…
  21. Mark Yatskar. 2019. A qualitative comparison of CoQA, SQuAD 2.0 and QuAC. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 2318–2323, Minneapolis, Minnesota. Asso… find
  22. Chen Zhu, Ankit Singh Rawat, Manzil Zaheer, Srinadh Bhojanapalli, Daliang Li, Felix Yu, and Sanjiv Kumar. 2020. Modifying memories in transformer models. arXiv preprint arXiv:2012.00363. find
  23. Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. SQuAD: 100,000+ questions for machine comprehension of text. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2383–2392, Austin, Texas. Association for Computational Linguistics.
  24. Hannah Rashkin, Vitaly Nikolaev, Matthew Lamm, Michael Collins, Dipanjan Das, Slav Petrov, Gaurav Singh Tomar, Iulia Turc, and David Reitter. 2021. Measuring attribution in natural language generation models. arXiv preprint arXiv:2112.12870.
  25. Adam Roberts, Hyung Won Chung, Anselm Levskaya, Gaurav Mishra, James Bradbury, Daniel Andor, Sharan Narang, Brian Lester, Colin Gaffney, Afroz Mohiuddin, et al. 2022. Scaling up models and data with t5x and seqio. arXiv preprint arXiv:2203.17189. find
  26. Adam Roberts, Colin Raffel, and Noam Shazeer. 2020. How much knowledge can you pack into the parameters of a language model? In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5418–5426, Online. Association for Computational Linguistics. find
  27. Elior Sulem, Jamaal Hay, and Dan Roth. 2021. Do we know what we don’t know? studying unanswerable questions beyond SQuAD 2.0. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 4543–4548, Punta Cana, Dominican Republic. Association for Computational Linguistics. find
  28. Neeraj Varshney, Swaroop Mishra, and Chitta Baral. 2022. Investigating selective prediction approaches across several tasks in IID, OOD, and adversarial settings. In Findings of the Association for Computational Linguistics: ACL 2022, pages 1995–2002, Dublin, Ireland. Association for Computational L… find
  29. Pat Verga, Haitian Sun, Livio Baldini Soares, and William Cohen. 2021. Adaptable and interpretable find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/neeman-2023-disentqa-disentangling-parametric-contextual-knowledge/