# Training-Free Cultural Alignment of Large Language Models via Persona Disagreement

arXiv: https://arxiv.org/abs/2605.10843 · HTML: https://mlsanity.com/paper/2605.10843/

## Key facts

- No citing papers in the map yet (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 26 of its 43 references that are in the map.
- First posted on arXiv 11 May 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, 26 of 26)

- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning](https://mlsanity.com/paper/1506.02142/) (2015)
- [Improving Factuality and Reasoning in Language Models through Multiagent Debate](https://mlsanity.com/paper/2305.14325/) (2023)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate](https://mlsanity.com/paper/2305.19118/) (2023)
- [Whose Opinions Do Language Models Reflect?](https://mlsanity.com/paper/2303.17548/) (2023)
- [Reinforcement Learning and Control as Probabilistic Inference: Tutorial and Review](https://mlsanity.com/paper/1805.00909/) (2018)
- [A Roadmap to Pluralistic Alignment](https://mlsanity.com/paper/2402.05070/) (2024)
- [BLEnD: A Benchmark for LLMs on Everyday Knowledge in Diverse Cultures and Languages](https://mlsanity.com/paper/2406.09948/) (2024)
- [The PRISM Alignment Dataset: What Participatory, Representative and Individualised Human Feedback Reveals About the Subjective and Multicultural Alignment of Large Language Models](https://mlsanity.com/paper/2404.16019/) (2024)
- [Controlled Decoding from Language Models](https://mlsanity.com/paper/2310.17022/) (2023)
- [ARGS: Alignment as Reward-Guided Search](https://mlsanity.com/paper/2402.01694/) (2024)
- [Information Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving](https://mlsanity.com/paper/1707.02342/) (2017)
- [Unintended Impacts of LLM Alignment on Global Representation](https://mlsanity.com/paper/2402.15018/) (2024)
- [Randomness, Not Representation: The Unreliability of Evaluating Cultural Alignment in LLMs](https://mlsanity.com/paper/2503.08688/) (2025)
- [Language Model Alignment in Multilingual Trolley Problems](https://mlsanity.com/paper/2407.02273/) (2024)
- [The Moral Machine Experiment on Large Language Models](https://mlsanity.com/paper/2309.05958/) (2023)
- [Large-scale moral machine experiment on large language models](https://mlsanity.com/paper/2411.06790/) (2024)
- [Exploring Persona-dependent LLM Alignment for the Moral Machine Experiment](https://mlsanity.com/paper/2504.10886/) (2025)
- [No Free Lunch in Language Model Bias Mitigation? Targeted Bias Reduction Can Exacerbate Unmitigated LLM Biases](https://mlsanity.com/paper/2511.18635/) (2025)
- [CAReDiO: Cultural Alignment via Representativeness and Distinctiveness Guided Data Optimization](https://mlsanity.com/paper/2504.08820/) (2025)
- [Culturally Grounded Personas in Large Language Models: Characterization and Alignment with Socio-Psychological Value Frameworks](https://mlsanity.com/paper/2601.22396/) (2026)
- [Mind the Gap in Cultural Alignment: Task-Aware Culture Management for Large Language Models](https://mlsanity.com/paper/2602.22475/) (2026)
- [Dropouts in Confidence: Moral Uncertainty in Human-LLM Alignment](https://mlsanity.com/paper/2511.13290/) (2025)

## Abstract (arXiv)

Large language models increasingly mediate decisions that turn on moral judgement, yet a growing body of evidence shows that their implicit preferences are not culturally neutral. Existing cultural alignment methods either require per-country preference data and fine-tuning budgets or assume white-box access to model internals that commercial APIs do not expose. In this work, we focus on this realistic black-box, public-data-only regime and observe that within-country sociodemographic disagreement, not consensus, is the primary steering signal. We introduce DISCA (Disagreement-Informed Steering for Cultural Alignment), an inference-time method that instantiates each country as a panel of World-Values-Survey-grounded persona agents and converts their disagreement into a bounded, loss-averse logit correction. Across 20 countries and 7 open-weight backbones (2B--70B), DISCA reduces cultural misalignment on MultiTP by 10--24% on the six backbones >=3.8B, and 2--7% on open-ended scenarios, without changing any weights. Our results suggest that inference-time calibration is a scalable alternative to fine-tuning for serving the long tail of global moral preferences.