Elamur Research⌕
☰
home / papers / 2026 / Beyond Marginal Distributions: A Framework to Evaluate the Representativeness of Demographic-Aligned LLMs

Beyond Marginal Distributions: A Framework to Evaluate the Representativeness of Demographic-Aligned LLMs

Paper · 2026 · Tristan Williams, Franziska Weeber, Sebastian Padó et al.

Summary

Studies that steer LLMs toward specific demographic groups usually judge success one survey question at a time, by comparing answer distributions. The authors argue this is too weak a test. Real populations also show consistent links between opinions across questions, and these links are central to cultural values theories. They propose an evaluation framework that measures both per-question (marginal) dissimilarity and how well a model reproduces inter-item correlation structures across subgroups.

As a case study, they use the World Values Survey (WVS) as ground truth and compare two steering methods:

  • demographic fine-tuning, using OpinionGPT's Reddit-trained adapters;
  • persona prompting of Phi-3-Mini-Instruct-4k.

OpinionGPT matches marginal distributions and variance better. Persona prompting does slightly better on correlation structure. Neither method comes close to the human correlation patterns.

The main lesson is that a model can look representative on marginals while failing structurally. Evaluations based only on marginals can therefore give an overly optimistic picture of demographic alignment.

Key points

  • The framework has two parts: averaged marginal dissimilarity over the question set Q, and a comparison of correlation matrices built from human and simulated responses. Responses can be aggregated by demographic dimension or by question topic.
  • Ten demographic subgroups are evaluated, matching OpinionGPT's adapters. They cover gender, US-style left/right politics, age groups (none under 30) and geographic regions or countries. OpinionGPT is built on Phi 3 Mini 4k.
  • The models generate 500 independent samples per survey item for each configuration. All prompting is in English only.
  • The ranking of the two methods reverses: fine-tuning wins on marginals, while persona prompting is marginally better on correlations.
  • Code is available at github.com/tdw75/beyond-marginal-distributions.

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

🎓University of StuttgartStuttgart · Germany94 works · 46,609 citations🎓Humboldt-Universität zu BerlinBerlin · Germany20 works · 41,895 citations

 Fields, topics and concepts

📚 Shelves 6The 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 worksSocial Sciences · 1,954 worksSafety Research · 626 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,894 works

Cite

Cite as Tristan Williams, Franziska Weeber, Sebastian Padó and Alan Akbik. 2026. Beyond Marginal Distributions: A Framework to Evaluate the Representativeness of Demographic-Aligned LLMs. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.236

@inproceedings{williams2026beyond,
  title = {{Beyond Marginal Distributions: A Framework to Evaluate the Representativeness of Demographic-Aligned LLMs}},
  author = {Tristan Williams, Franziska Weeber, Sebastian Padó and Alan Akbik},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.236},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.236}
}

References

43 references, 17 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, Hany Awadalla, Ahmed Awadallah, Ammar Ahmad Awan, Nguyen Bach, Amit Bahree, Arash Bakhtiari, Jianmin Bao, Harkirat Behl, Alon Benhaim, Misha Bilenko, Johan Bjorck, Sébastien Bubeck, Martin Cai, Qin Cai, Vishrav Chaudhary, Dong Chen, Dongdong Chen, and 110 others. 2024. Phi-… — Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone
  2. Lisa P. Argyle, Ethan C. Busby, Nancy Fulda, Joshua R. Gubler, Christopher Rytting, and David Wingate. 2023. Out of One, Many: Using Language Models to Simulate Human Samples. Political Analysis, 31(3):337–351.
  3. James Bisbee, Joshua D. Clinton, Cassy Dorff, Brenton Kenkel, and Jennifer M. Larson. 2024. Synthetic Replacements for Human Survey Data? The Perils of Large Language Models. Political Analysis, 32(4):401–416. find
  4. Tanise Ceron, Neele Falk, Ana Barić, Dmitry Nikolaev, and Sebastian Padó. 2024. Beyond Prompt Brittleness: Evaluating the Reliability and Consistency of Political Worldviews in LLMs. Transactions of the Association for Computational Linguistics, 12:1378– 1400.
  5. Ricardo Dominguez-Olmedo, Moritz Hardt, and Celestine Mendler-Dünner. 2024. Questioning the Survey Responses of Large Language Models. Advances in Neural Information Processing Systems, pages 45850–45878. find
  6. Esin Durmus, Karina Nguyen, Thomas Liao, Nicholas Schiefer, Amanda Askell, Anton Bakhtin, Carol Chen, Zac Hatfield-Dodds, Danny Hernandez, Nicholas Joseph, Liane Lovitt, Sam McCandlish, Orowa Sikder, Alex Tamkin, Janel Thamkul, Jared Kaplan, Jack Clark, and Deep Ganguli. 2024. Towards Measuring the…
  7. Shangbin Feng, Chan Young Park, Yuhan Liu, and Yulia Tsvetkov. 2023. From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume…
  8. Sasuke Fujimoto and Kazuhiro Takemoto. 2023. Revisiting the political biases of ChatGPT. Frontiers in Artificial Intelligence, 6:1232003. find
  9. Iason Gabriel. 2020. Artificial Intelligence, Values, and Alignment. Minds and Machines, 30(3):411–437. find
  10. Jamie Lynn Goodwin, Andrew Lloyd Williams, and Patricia Snell Herzog. 2020. Cross-Cultural Values: A Meta-Analysis of Major Quantitative Studies in the Last Decade (2010–2020). Religions, 11(8):396. Multidisciplinary Digital Publishing Institute. find
  11. Suriya Gunasekar, Yi Zhang, Jyoti Aneja, Caio César Teodoro Mendes, Allie Del Giorno, Sivakanth Gopi, Mojan Javaheripi, Piero Kauffmann, Gustavo de Rosa, Olli Saarikivi, Adil Salim, Shital Shah, Harkirat Singh Behl, Xin Wang, Sébastien Bubeck, Ronen Eldan, Adam Tauman Kalai, Yin Tat Lee, and Yuanzhi… find
  12. Shashank Gupta, Vaishnavi Shrivastava, Ameet Deshpande, Ashwin Kalyan, Peter Clark, Ashish Sabharwal, and Tushar Khot. 2024. Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs. In The Twelfth International Conference on Learning Representations. ArXiv:2311.04892 [cs].
  13. Christian Haerpfer, Ronald Inglehart, Alejandro Moreno, Christian Welzel, Kseniya Kizilova, Jaime Diez-Medrano, Marta Lagos, Pippa Norris, Eduard Ponarin, and Bi Puranen. 2020. World Values Survey Wave 7 (2017-2020) Cross-National Data-Set. find
  14. Patrick Haller, Ansar Aynetdinov, and Alan Akbik. 2024. OpinionGPT: Modelling Explicit Biases in Instruction-Tuned LLMs. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 3: System Demonstrations)… find
  15. Jochen Hartmann, Jasper Schwenzow, and Maximilian Witte. 2023. The political ideology of conversational AI: Converging evidence on ChatGPT’s pro-environmental, left-libertarian orientation. arXiv preprint. ArXiv:2301.01768 [cs].
  16. Geert Hofstede. 1980. Culture’s Consequences: International Differences in Work-Related Values. Sage. find
  17. Allyson L. Holbrook, Jon A. Krosnick, David Moore, and Roger Tourangeau. 2007. Response Order Effects in Dichotomous Categorical Questions Presented Orally: The Impact of Question and Respondent Attributes. The Public Opinion Quarterly, 71(3):325–348. Publisher: [Oxford University Press, American As… find
  18. Geoff Hollis, Chris Westbury, and Lianne Lefsrud. 2017. Extrapolating human judgments from skip-gram vector representations of word meaning. Quarterly Journal of Experimental Psychology, 70(8):1603–1619. find
  19. Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. 2020. The Curious Case of Neural Text Degeneration. In 8th International Conference on Learning Representations (ICLR 2020). ArXiv:1904.09751 [cs].
  20. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models. In 10th International Conference on Learning Representations.
  21. Ronald Inglehart. 2000. Globalization and postmodern values. The Washington Quarterly, 23(1):215–228. find
  22. Ronald Inglehart and Christian Welzel. 2023. The Inglehart-Welzel World Cultural Map. World Values Survey 7. find
  23. Shariar Kabir, Kevin Esterling, and Yue Dong. 2025. Testing Conviction: An Argumentative Framework for Measuring LLM Political Stability. arXiv preprint. ArXiv:2504.17052 [cs]. — [arXiv:2504.17052](https://arxiv.org/abs/2504.17052) find
  24. Robert Kirk, Ishita Mediratta, Christoforos Nalmpantis, Jelena Luketina, Eric Hambro, Edward Grefenstette, and Roberta Raileanu. 2024. Understanding the Effects of RLHF on LLM Generalisation and Diversity. In 12th International Conference on Learning Representations.
  25. Jon A. Krosnick and Duane F. Alwin. 1987. An Evaluation of a Cognitive Theory of Response-Order Effects in Survey Measurement. Public Opinion Quarterly, 51(2):201. find
  26. Alina Leidinger, Robert Van Rooij, and Ekaterina Shutova. 2023. The language of prompting: What linguistic properties make a prompt successful? In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9210–9232, Singapore. Association for Computational Linguistics. find
  27. Cheng Li, Mengzhuo Chen, Jindong Wang, Sunayana Sitaram, and Xing Xie. 2025. CultureLLM: incorporating cultural differences into large language models. In Proceedings of the 38th International Conference on Neural Information Processing Systems, pages 84799–84838, Red Hook, NY, USA. Curran Associate…
  28. Percy Liang, Rishi Bommasani, Tony Lee, Dimitris Tsipras, Dilara Soylu, Michihiro Yasunaga, Yian Zhang, Deepak Narayanan, Yuhuai Wu, Ananya Kumar, Benjamin New-Man, Binhang Yuan, Bobby Yan, Ce Zhang, Christian Cosgrove, Christopher D. Manning, Christopher Ré, Diana Acosta-Navas, Drew A. Hudson, and… find
  29. Christina Lu and Max Van Kleek. 2024. Model Plurality: A Taxonomy for Pluralistic AI. In Pluralistic Alignment Workshop at NeurIPS 2024. find
  30. John Levi Martin. 2023. The Ethico-Political Universe of ChatGPT. Journal of Social Computing, 4(1):1– 11. find
  31. Rudra Murthy, Prince Kumar, Praveen Venkateswaran, and Danish Contractor. 2024. Evaluating the Instruction-following Abilities of Language Models using Knowledge Tasks. arXiv preprint. ArXiv:2410.12972 [cs] version: 1. — [arXiv:2410.12972](https://arxiv.org/abs/2410.12972) find
  32. Simon Münker. 2025. Fingerprinting LLMs through Survey Item Factor Correlation: A Case Study on Humor Style Questionnaire. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 245–258, Suzhou, China. Association for Computational Linguistics. find
  33. Ethan Perez, Sam Ringer, Kamile Lukosiute, Karina Nguyen, Edwin Chen, Scott Heiner, Craig Pettit, Catherine Olsson, Sandipan Kundu, Saurav Kadavath, Andy Jones, Anna Chen, Benjamin Mann, Brian Israel, Bryan Seethor, Cameron McKinnon, Christopher Olah, Da Yan, Daniela Amodei, and 44 others. 2023. Dis… — Discovering Language Model Behaviors with Model-Written Evaluations
  34. Pew Research Center. 2024. Social Media Fact Sheet. find
  35. Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, and Tatsunori Hashimoto. 2023. Whose Opinions Do Language Models Reflect? In Proceedings of the 40th International Conference on Machine Learning, pages 29971–30004. PMLR. ISSN: 2640-3498.
  36. Shalom H. Schwartz. 1992. Universals in the Content and Structure of Values: Theoretical Advances and Empirical Tests in 20 Countries. In Advances in Experimental Social Psychology, volume 25, pages 1–65. Elsevier. find
  37. Indira Sen, Marlene Lutz, Elisa Rogers, David Garcia, and Markus Strohmaier. 2025. Missing the Margins: A Systematic Literature Review on the Demographic Representativeness of LLMs. In Findings of the Association for Computational Linguistics: ACL 2025, pages 24263–24289, Vienna, Austria. Associatio… find
  38. Taylor Sorensen, Jared Moore, Jillian Fisher, Mitchell Gordon, Niloofar Mireshghallah, Christopher Michael Rytting, Andre Ye, Liwei Jiang, Ximing Lu, Nouha Dziri, Tim Althoff, and Yejin Choi. 2024. Position: a roadmap to pluralistic alignment. In Proceedings of the 41st International Conference on M…
  39. Joseph Suh, Erfan Jahanparast, Suhong Moon, Minwoo Kang, and Serina Chang. 2025. Language Model Fine-Tuning on Scaled Survey Data for Predicting Distributions of Public Opinions. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages…
  40. Yan Tao, Olga Viberg, Ryan S Baker, and René F Kizilcec. 2024. Cultural Bias and Cultural Alignment of Large Language Models. PNAS Nexus, 3(9):pgae346.
  41. Cédric Villani. 2009. The Wasserstein distances. In Cédric Villani, editor, Optimal Transport: Old and New, pages 93–111. Springer, Berlin, Heidelberg. find
  42. Franziska Weeber, Tanise Ceron, and Sebastian Padó. 2026. Do political opinions transfer between western languages? an analysis of unaligned and aligned multilingual LLMs. In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long P…
  43. Laura Weidinger, Jonathan Uesato, Maribeth Rauh, Conor Griffin, Po-Sen Huang, John Mellor, Amelia Glaese, Myra Cheng, Borja Balle, Atoosa Kasirzadeh, Courtney Biles, Sasha Brown, Zac Kenton, Will Hawkins, Tom Stepleton, Abeba Birhane, Lisa Anne Hendricks, Laura Rimell, William Isaac, and 4 others. 2… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/williams-2026-beyond-marginal-distributions-framework-evaluate/