As Large Language Models (LLMs) gain stronger multilingual capabilities, their ability to handle culturally diverse entities becomes crucial. Prior work has shown that LLMs often favor Western-associated entities in Arabic, raising concerns about cultural fairness. Due to the lack of multilingual benchmarks, it remains unclear if such biases also manifest in different non-Western languages. In this paper, we introduce Camellia, a benchmark for measuring entity-centric cultural biases in nine Asian languages spanning six distinct Asian cultures. Camellia includes 19,530 entities manually annotated for association with the specific Asian or Western culture, as well as 2,173 naturally occurring masked contexts for entities derived from social media posts. Using Camellia, we evaluate cultural biases in four recent multilingual LLM families across various tasks such as cultural context adaptation, sentiment association, and entity extractive QA. Our analyses show a struggle by LLMs at cultural adaptation in all Asian languages, with performance differing across models developed in regions with varying access to culturally-relevant data. We further observe that different LLM families hold their distinct biases, differing in how they associate cultures with particular sentiments. Lastly, we find that LLMs struggle with context understanding in Asian languages, creating performance gaps between cultures in entity extraction. Large Language Models (LLMs) have rapidly integrated into modern technology, serving users from diverse cultures (Adilazuarda et al., 2024). Among the vast range of text they process, LLMs frequently encounter entities such as people's names, locations, or food dishes, which are pervasive in text corpora (Wolfe & Caliskan, 2021; Pawar et al., 2025a) and often appear in user prompts (Li et al., 2024a; Wang et al., 2025). Importantly, entities carry cultural associations, making it essential for LLMs to handle culturally diverse entities fairly.

As large language models (LLMs) are increasingly deployed in diverse cultural environments, evaluating their cultural understanding capability has become essential for ensuring trustworthy and culturally aligned applications. However, most existing benchmarks lack comprehensiveness and are challenging to scale and adapt across different cultural contexts, because their frameworks often lack guidance from well-established cultural theories and tend to rely on expert-driven manual annotations. To address these issues, we propose CultureScope, the most comprehensive evaluation framework to date for assessing cultural understanding in LLMs. Inspired by the cultural iceberg theory, we design a novel dimensional schema for cultural knowledge classification, comprising 3 layers and 140 dimensions, which guides the automated construction of culture-specific knowledge bases and corresponding evaluation datasets for any given languages and cultures. Experimental results demonstrate that our method can effectively evaluate cultural understanding. They also reveal that existing large language models lack comprehensive cultural competence, and merely incorporating multilingual data does not necessarily enhance cultural understanding.

Large language models exhibit cultural biases and limited cross-cultural understanding capabilities, particularly when serving diverse global user populations. We propose MCEval, a novel multilingual evaluation framework that employs dynamic cultural question construction and enables causal analysis through Counterfactual Rephrasing and Confounder Rephrasing. Our comprehensive evaluation spans 13 cultures and 13 languages, systematically assessing both cultural awareness and cultural bias across different linguistic scenarios. The framework provides 39,897 cultural awareness instances and 17,940 cultural bias instances. Experimental results reveal performance disparities across different linguistic scenarios, demonstrating that optimal cultural performance is not only linked to training data distribution, but also is related to language-culture alignment. The evaluation results also expose the fairness issue, where approaches appearing successful in the English scenario create substantial disadvantages. MCEval represents the first comprehensive multilingual cultural evaluation framework that provides deeper insights into LLMs' cultural understanding.

Artificial intelligences picking up sexist and racist biases is a well-known and persistent problem, but researchers are now turning this to their advantage to analyse social attitudes through history. Training AI models on novels from a certain decade can instil them with the prejudices of that era, offering a new way to study how cultural biases have evolved over time. Large language models (LLMs) such as ChatGPT learn by analysing large collections of text.

As large language models (LLMs) like GPT-4 and Llama 3 become integral to educational contexts, concerns are mounting over the cultural biases, power imbalances, and ethical limitations embedded within these technologies. Though generative AI tools aim to enhance learning experiences, they often reflect values rooted in Western, Educated, Industrialized, Rich, and Democratic (WEIRD) cultural paradigms, potentially sidelining diverse global perspectives. This paper proposes a framework to assess and mitigate cultural bias within LLMs through the lens of applied multiplexity. Multiplexity, inspired by Senturk et al. and rooted in Islamic and other wisdom traditions, emphasizes the coexistence of diverse cultural viewpoints, supporting a multi-layered epistemology that integrates both empirical sciences and normative values. Our analysis reveals that LLMs frequently exhibit cultural polarization, with biases appearing in both overt responses and subtle contextual cues. To address inherent biases and incorporate multiplexity in LLMs, we propose two strategies: \textit{Contextually-Implemented Multiplex LLMs}, which embed multiplex principles directly into the system prompt, influencing LLM outputs at a foundational level and independent of individual prompts, and \textit{Multi-Agent System (MAS)-Implemented Multiplex LLMs}, where multiple LLM agents, each representing distinct cultural viewpoints, collaboratively generate a balanced, synthesized response. Our findings demonstrate that as mitigation strategies evolve from contextual prompting to MAS-implementation, cultural inclusivity markedly improves, evidenced by a significant rise in the Perspectives Distribution Score (PDS) and a PDS Entropy increase from 3.25\% at baseline to 98\% with the MAS-Implemented Multiplex LLMs. Sentiment analysis further shows a shift towards positive sentiment across cultures,...

Cultural values alignment in Large Language Models (LLMs) is a critical challenge due to their tendency to embed Western-centric biases from training data, leading to misrepresentations and fairness issues in cross-cultural contexts. Recent approaches, such as role-assignment and few-shot learning, often struggle with reliable cultural alignment as they heavily rely on pre-trained knowledge, lack scalability, and fail to capture nuanced cultural values effectively. To address these issues, we propose ValuesRAG, a novel and effective framework that applies Retrieval-Augmented Generation (RAG) with in-context learning to integrate cultural and demographic knowledge dynamically during text generation. Leveraging the World Values Survey (WVS) dataset, ValuesRAG first generates summaries of values for each individual. Subsequently, we curated several representative regional datasets to serve as test datasets and retrieve relevant summaries of values based on demographic features, followed by a reranking step to select the top-k relevant summaries. ValuesRAG consistently outperforms baseline methods, both in the main experiment and in the ablation study where only the values summary was provided, highlighting ValuesRAG's potential to foster culturally aligned AI systems and enhance the inclusivity of AI-driven applications.

Large Language Model (LLM)-based recommendation systems provide more comprehensive recommendations than traditional systems by deeply analyzing content and user behavior. However, these systems often exhibit biases, favoring mainstream content while marginalizing non-traditional options due to skewed training data. This study investigates the intricate relationship between bias and LLM-based recommendation systems, with a focus on music, song, and book recommendations across diverse demographic and cultural groups. Through a comprehensive analysis conducted over different LLM-models, this paper evaluates the impact of bias on recommendation outcomes. Our findings reveal that bias is so deeply ingrained within these systems that even a simpler intervention like prompt engineering can significantly reduce bias, underscoring the pervasive nature of the issue. Moreover, factors like intersecting identities and contextual information, such as socioeconomic status, further amplify these biases, demonstrating the complexity and depth of the challenges faced in creating fair recommendations across different groups.

Popularity bias in music recommendation systems -- where artists and tracks with the highest listen counts are recommended more often -- can also propagate biases along demographic and cultural axes. In this work, we identify these biases in recommendations for artists from underrepresented cultural groups in prototype-based matrix factorization methods. Unlike traditional matrix factorization methods, prototype-based approaches are interpretable. This allows us to directly link the observed bias in recommendations for minority artists (the effect) to specific properties of the embedding space (the cause). We mitigate popularity bias in music recommendation through capturing both users' and songs' cultural nuances in the embedding space. To address these challenges while maintaining recommendation quality, we propose two novel enhancements to the embedding space: i) we propose an approach to filter-out the irrelevant prototypes used to represent each user and item to improve generalizability, and ii) we introduce regularization techniques to reinforce a more uniform distribution of prototypes within the embedding space. Our results demonstrate significant improvements in reducing popularity bias and enhancing demographic and cultural fairness in music recommendations while achieving competitive -- if not better -- overall performance.

Training LLMs in low resources languages usually utilizes data augmentation with machine translation (MT) from English language. However, translation brings a number of challenges: there are large costs attached to translating and curating huge amounts of content with high-end machine translation solutions, the translated content carries over cultural biases, and if the translation is not faithful and accurate, the quality of the data degrades causing issues in the trained model. In this work we investigate the role of translation and synthetic data in training language models. We translate TinyStories, a dataset of 2.2M short stories for 3-4 year old children, from English to Arabic using the free NLLB-3B MT model. We train a number of story generation models of sizes 1M-33M parameters using this data. We identify a number of quality and task-specific issues in the resulting models. To rectify these issues, we further pre-train the models with a small dataset of synthesized high-quality stories, representing 1\% of the original training data, using a capable LLM in Arabic. We show using GPT-4 as a judge and dictionary learning analysis from mechanistic interpretability that the suggested approach is a practical means to resolve some of the translation pitfalls. We illustrate the improvement through case studies of linguistic issues and cultural bias.

Given the rapid advancement of large-scale language models, artificial intelligence (AI) models, like ChatGPT, are playing an increasingly prominent role in human society. However, to ensure that artificial intelligence models benefit human society, we must first fully understand the similarities and differences between the human-like characteristics exhibited by artificial intelligence models and real humans, as well as the cultural stereotypes and biases that artificial intelligence models may exhibit in the process of interacting with humans. This study first measured ChatGPT in 84 dimensions of psychological characteristics, revealing differences between ChatGPT and human norms in most dimensions as well as in high-dimensional psychological representations. Additionally, through the measurement of ChatGPT in 13 dimensions of cultural values, it was revealed that ChatGPT's cultural value patterns are dissimilar to those of various countries/regions worldwide. Finally, an analysis of ChatGPT's performance in eight decision-making tasks involving interactions with humans from different countries/regions revealed that ChatGPT exhibits clear cultural stereotypes in most decision-making tasks and shows significant cultural bias in third-party punishment and ultimatum games. The findings indicate that, compared to humans, ChatGPT exhibits a distinct psychological profile and cultural value orientation, and it also shows cultural biases and stereotypes in interpersonal decision-making. Future research endeavors should emphasize enhanced technical oversight and augmented transparency in the database and algorithmic training procedures to foster more efficient cross-cultural communication and mitigate social disparities.