# Steering LLMs for Culturally Localized Generation

arXiv: https://arxiv.org/abs/2603.23301 · HTML: https://mlsanity.com/paper/2603.23301/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Knowledge and representations inside LLMs, one of 13 directions in Mechanistic interpretability.
- Builds on 32 of its 34 references that are in the map.
- First posted on arXiv 24 Mar 2026.

## Placed in

- Mechanistic interpretability / [Knowledge and representations inside LLMs](https://mlsanity.com/mechinterp/knowledge-and-representations-inside-llms/)

## Builds on (most cited first, 30 of 32)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Datasheets for Datasets](https://mlsanity.com/paper/1803.09010/) (2018)
- [Language (Technology) is Power: A Critical Survey of "Bias" in NLP](https://mlsanity.com/paper/2005.14050/) (2020)
- [How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings](https://mlsanity.com/paper/1909.00512/) (2019)
- [Whose Opinions Do Language Models Reflect?](https://mlsanity.com/paper/2303.17548/) (2023)
- [Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2307.15217/) (2023)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [Marked Personas: Using Natural Language Prompts to Measure Stereotypes in Language Models](https://mlsanity.com/paper/2305.18189/) (2023)
- [Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs](https://mlsanity.com/paper/2311.04892/) (2023)
- [Towards Measuring and Modeling "Culture" in LLMs: A Survey](https://mlsanity.com/paper/2403.15412/) (2024)
- [Personalizing Reinforcement Learning from Human Feedback with Variational Preference Learning](https://mlsanity.com/paper/2408.10075/) (2024)
- [Cultural Bias and Cultural Alignment of Large Language Models](https://mlsanity.com/paper/2311.14096/) (2023)
- [Investigating Cultural Alignment of Large Language Models](https://mlsanity.com/paper/2402.13231/) (2024)
- [Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoders](https://mlsanity.com/paper/2410.20526/) (2024)
- [SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability](https://mlsanity.com/paper/2503.09532/) (2025)
- [MaxMin-RLHF: Alignment with Diverse Human Preferences](https://mlsanity.com/paper/2402.08925/) (2024)
- [Extracting Cultural Commonsense Knowledge at Scale](https://mlsanity.com/paper/2210.07763/) (2022)
- [On the Algorithmic Bias of Aligning Large Language Models with RLHF: Preference Collapse and Matching Regularization](https://mlsanity.com/paper/2405.16455/) (2024)
- [Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2504.12501/) (2025)
- [Culture is Not Trivia: Sociocultural Theory for Cultural NLP](https://mlsanity.com/paper/2502.12057/) (2025)
- [Culturally Aware and Adapted NLP: A Taxonomy and a Survey of the State of the Art](https://mlsanity.com/paper/2406.03930/) (2024)
- [Route Sparse Autoencoder to Interpret Large Language Models](https://mlsanity.com/paper/2503.08200/) (2025)
- [Multilingual Prompting for Improving LLM Generation Diversity](https://mlsanity.com/paper/2505.15229/) (2025)
- [Global Voices, Local Biases: Socio-Cultural Prejudices across Languages](https://mlsanity.com/paper/2310.17586/) (2023)
- [Localized Cultural Knowledge is Conserved and Controllable in Large Language Models](https://mlsanity.com/paper/2504.10191/) (2025)
- [Ethics Sheets for AI Tasks](https://mlsanity.com/paper/2107.01183/) (2021)
- [Translating Across Cultures: LLMs for Intralingual Cultural Adaptation](https://mlsanity.com/paper/2406.14504/) (2024)
- [Internal Value Alignment in Large Language Models through Controlled Value Vector Activation](https://mlsanity.com/paper/2507.11316/) (2025)

## Built on by (most cited first, 4 of 4)

- [Personalization, Personas, and Forecasting in Value Alignment](https://mlsanity.com/paper/2607.24782/) (2026)
- [Cultural Binding Heads in Language Models](https://mlsanity.com/paper/2605.28543/) (2026)
- [Right Frame, Wrong Rule: Cultural Cues Expose the Financial Knowledge Gap They Were Meant to Close](https://mlsanity.com/paper/2609.00999/) (2026)
- [Geometric Representations of African Languages: A Regional Semantic Hub and Cultural Steering](https://mlsanity.com/paper/2609.36205/) (2026)

## Abstract (arXiv)

LLMs are deployed globally, yet produce responses biased towards cultures with abundant training data. Existing cultural localization approaches such as prompting or post-training alignment are black-box, hard to control, and do not reveal whether failures reflect missing knowledge or poor elicitation. In this paper, we address these gaps using mechanistic interpretability to uncover and manipulate cultural representations in LLMs. Leveraging sparse autoencoders, we identify interpretable features that encode culturally salient information and aggregate them into Cultural Embeddings (CuE). We use CuE both to analyze implicit cultural biases under underspecified prompts and to construct white-box steering interventions. Across multiple models, we show that CuE-based steering increases cultural faithfulness and elicits significantly rarer, long-tail cultural concepts than prompting alone. Notably, CuE-based steering is complementary to black-box localization methods, offering gains when applied on top of prompt-augmented inputs. This also suggests that models do benefit from better elicitation strategies, and don't necessarily lack long-tail knowledge representation, though this varies across cultures. Our results provide both diagnostic insight into cultural representations in LLMs and a controllable method to steer towards desired cultures.