# The Limits of Simulated Societies: How Post-Training and Survey Fine-Tuning Erase Cross-Cultural Variance

arXiv: https://arxiv.org/abs/2609.25760 · HTML: https://mlsanity.com/paper/2609.25760/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Human values and AI feedback alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 14 of its 23 references that are in the map.
- First posted on arXiv 22 Sep 2026.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Human values and AI feedback alignment](https://mlsanity.com/alignment/human-values-and-ai-feedback-alignment/)

## Builds on (most cited first, 14 of 14)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [Whose Opinions Do Language Models Reflect?](https://mlsanity.com/paper/2303.17548/) (2023)
- [Out of One, Many: Using Language Models to Simulate Human Samples](https://mlsanity.com/paper/2209.06899/) (2022)
- [Towards Measuring the Representation of Subjective Global Opinions in Language Models](https://mlsanity.com/paper/2306.16388/) (2023)
- [Large language models that replace human participants can harmfully misportray and flatten identity groups](https://mlsanity.com/paper/2402.01908/) (2024)
- [Random Silicon Sampling: Simulating Human Sub-Population Opinion Using a Large Language Model Based on Group-Level Demographic Information](https://mlsanity.com/paper/2402.18144/) (2024)
- [AI-Augmented Surveys: Leveraging Large Language Models and Surveys for Opinion Prediction](https://mlsanity.com/paper/2305.09620/) (2023)
- [Finetuning LLMs for Human Behavior Prediction in Social Science Experiments](https://mlsanity.com/paper/2509.05830/) (2025)
- [Valid Survey Simulations with Limited Human Data: The Roles of Prompting, Fine-Tuning, and Rectification](https://mlsanity.com/paper/2510.11408/) (2025)
- [Assessing Generalization for Subpopulation Representative Modeling via In-Context Learning](https://mlsanity.com/paper/2402.07368/) (2024)

## Built on by (most cited first, 1 of 1)

- [When Should an In-Context Learner Expand Its Hypothesis Space?](https://mlsanity.com/paper/2610.09471/) (2026)

## Abstract (arXiv)

Using large language models (LLMs) to simulate diverse human populations has the potential to transform many aspects of computational social science, yet many evaluations score the average response rather than the spread of opinion within real groups. Here, we develop a diagnostic framework that measures point accuracy alongside dispersion retention, the ratio of predicted to human standard deviation (dr), on 10,000 respondent--question pairs from the World Values Survey (WVS) spanning twelve countries and six continents. We evaluate eleven zero-shot language models and five variants fine-tuned on WVS data with SFT, DPO, and GRPO. We identify a failure mode we term textitconsensus collapse, where alignment training compresses outputs toward one stereotype per group. Along the post-training trajectory from the Llama~3.1 70B base to the Tulu~3 checkpoints, the first stage, supervised instruction tuning, removes half of the spread with minimal accuracy gain (dr 1.22 to 0.59; accuracy +0.9 points), the later stages do not restore it, and a gap opens between WEIRD and non-WEIRD countries that survey fine-tuning then deepens while pursuing higher point accuracy. The most accurate model (Tulu~3 70B-DPO fine-tuned on WVS, 57.9%) keeps half the human spread overall (dr = 0.50) and 11% of it for Nigeria, against 0.70--0.87 for WEIRD countries. Raising the sampling temperature to 1.0 leaves the Wasserstein-1 distance (wone) to human distributions unchanged for both fine-tuned DPO models, and GRPO on Qwen~3.5 9B does not restore the spread under either an accuracy reward or a distribution-shaped reward. Mixing the aligned model with an unaligned prior raises dr from 0.51 to 0.62 on a held-out split but leaves Nigeria at 0.36. Point accuracy alone therefore misjudges these simulators, and current post-training trades diversity for consensus.