StereoTales: A Multilingual Framework for Open-Ended Stereotype Discovery in LLMs

• URL: http://arxiv.org/abs/2605.10442v2
• Date: Tue, 12 May 2026 10:07:26 GMT
• Title: StereoTales: A Multilingual Framework for Open-Ended Stereotype Discovery in LLMs
• Abstract summary: StereoTales is a multilingual dataset and evaluation pipeline for studying the emergence of social bias in open-ended LLM generation.<n>The dataset covers 10 languages and 79 socio-demographic attributes, and comprises over 650k stories generated by 23 recent LLMs.<n>From these, we apply statistical tests to identify more than 1,500 over-represented associations, which we then rate for harmfulness through both a panel of humans and the same LLMs.
• Score: 2.5649028885424374
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multilingual studies of social bias in open-ended LLM generation remain limited: most existing benchmarks are English-centric, template-based, or restricted to recognizing pre-specified stereotypes. We introduce StereoTales, a multilingual dataset and evaluation pipeline for systematically studying the emergence of social bias in open-ended LLM generation. The dataset covers 10 languages and 79 socio-demographic attributes, and comprises over 650k stories generated by 23 recent LLMs, each annotated with the socio-demographic profile of the protagonist across 19 dimensions. From these, we apply statistical tests to identify more than 1{,}500 over-represented associations, which we then rate for harmfulness through both a panel of humans (N = 247) and the same LLMs. We report three main findings. \textbf{(i)} Every model we evaluate emits consequential harmful stereotypes in open-ended generation, regardless of size or capabilities, and these associations are largely shared across providers rather than isolated misbehaviors. \textbf{(ii)} Prompt language strongly shapes which stereotypes appear: rather than transferring as a shared set of biases, harmful associations adapt culturally to the prompt language and amplify bias against locally salient protected groups. \textbf{(iii)} Human and LLM harmfulness judgments are broadly aligned (Spearman $ρ=0.62$), with disagreements concentrating on specific attribute classes rather than specific providers. To support further analyses, we release the evaluation code and the dataset, including model generations, attribute annotations, and harmfulness ratings.

Related papers

• Scalable and Culturally Specific Stereotype Dataset Construction via Human-LLM Collaboration [36.09273812207085]
  EspanStereo is a Spanish-language stereotype dataset spanning multiple Spanish-speaking countries across Europe and Latin America.<n>Our framework is adaptable to other languages and regions, offering a scalable path toward multilingual stereotype benchmarks.<n>This work broadens the scope of stereotype analysis in large language models and lays the groundwork for comprehensive cross-cultural bias evaluation.
  arXiv  Detail & Related papers  (2026-07-08T20:06:00Z)
• Us-vs-Them bias in Large Language Models [0.569978892646475]
  We find consistent ingroup-positive and outgroup-negative associations across foundational large language models.<n>For personas examined, conservative personas exhibit greater outgroup hostility, whereas liberal personas display stronger ingroup solidarity.
  arXiv  Detail & Related papers  (2025-12-03T07:11:22Z)
• A Multilingual, Large-Scale Study of the Interplay between LLM Safeguards, Personalisation, and Disinformation [12.577461004484604]
  This study presents the first large-scale, multilingual analysis of persona-targeted disinformation generation by Large Language Models (LLMs)<n>We prompt eight state-of-the-art LLMs with 324 false narratives and 150 demographic personas across four languages--English, Russian, Portuguese, and Hindi--resulting in AI-TRAITS.<n>Results show that the use of even simple personalisation prompts significantly increases the likelihood of jailbreaks.
  arXiv  Detail & Related papers  (2025-10-14T21:10:50Z)
• Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models [81.45743826739054]
  A major barrier has been the lack of demographic annotations in web-scale datasets such as LAION-400M.<n>We create person-centric annotations for the full dataset, including over 276 million bounding boxes, perceived gender and race/ethnicity labels, and automatically generated captions.<n>Using them, we uncover demographic imbalances and harmful associations, such as the disproportionate linking of men and individuals perceived as Black or Middle Eastern with crime-related and negative content.
  arXiv  Detail & Related papers  (2025-10-04T07:51:59Z)
• Which Cultural Lens Do Models Adopt? On Cultural Positioning Bias and Agentic Mitigation in LLMs [53.07843733899881]
  Large language models (LLMs) have unlocked a wide range of downstream generative applications.<n>We find that they also risk perpetuating subtle fairness issues tied to culture, positioning their generations from the perspectives of the mainstream US culture.<n>We propose 2 inference-time mitigation methods to resolve these biases.
  arXiv  Detail & Related papers  (2025-09-25T12:28:25Z)
• Simulating Identity, Propagating Bias: Abstraction and Stereotypes in LLM-Generated Text [8.988787218035414]
  We investigate whether persona-prompting leads to different levels of linguistic abstraction when generating short texts linking socio-demographic categories with stereotypical or non-stereotypical attributes.<n>Our results confirm criticisms about the ecology of personas as representative of socio-demographic groups and raise concerns about the risk of propagating stereotypes.
  arXiv  Detail & Related papers  (2025-09-10T10:49:21Z)
• Discovering Bias Associations through Open-Ended LLM Generations [1.7373859011890633]
  Social biases embedded in Large Language Models (LLMs) raise critical concerns.<n>We present the Bias Association Discovery Framework (BADF), a systematic approach for extracting associations between demographic identities and descriptive concepts.<n>Our findings advance the understanding of biases in open-ended generation and provide a scalable tool for identifying and analyzing bias associations in LLMs.
  arXiv  Detail & Related papers  (2025-08-02T15:31:55Z)
• LIBRA: Measuring Bias of Large Language Model from a Local Context [9.612845616659776]
  Large Language Models (LLMs) have significantly advanced natural language processing applications.<n>Yet their widespread use raises concerns regarding inherent biases that may reduce utility or harm for particular social groups.<n>This research addresses these limitations with a Local Integrated Bias Recognition and Assessment Framework (LIBRA) for measuring bias.
  arXiv  Detail & Related papers  (2025-02-02T04:24:57Z)
• Evaluating Implicit Bias in Large Language Models by Attacking From a Psychometric Perspective [66.34066553400108]
  We conduct a rigorous evaluation of large language models' implicit bias towards certain demographics.<n>Inspired by psychometric principles, we propose three attack approaches, i.e., Disguise, Deception, and Teaching.<n>Our methods can elicit LLMs' inner bias more effectively than competitive baselines.
  arXiv  Detail & Related papers  (2024-06-20T06:42:08Z)
• White Men Lead, Black Women Help? Benchmarking and Mitigating Language Agency Social Biases in LLMs [58.27353205269664]
  Social biases can manifest in language agency in Large Language Model (LLM)-generated content.<n>We introduce the Language Agency Bias Evaluation benchmark, which comprehensively evaluates biases in LLMs.<n>Using LABE, we unveil language agency social biases in 3 recent LLMs: ChatGPT, Llama3, and Mistral.
  arXiv  Detail & Related papers  (2024-04-16T12:27:54Z)
• Social Bias Probing: Fairness Benchmarking for Language Models [38.180696489079985]
  This paper proposes a novel framework for probing language models for social biases by assessing disparate treatment. We curate SoFa, a large-scale benchmark designed to address the limitations of existing fairness collections. We show that biases within language models are more nuanced than acknowledged, indicating a broader scope of encoded biases than previously recognized.
  arXiv  Detail & Related papers  (2023-11-15T16:35:59Z)
• Bias and Fairness in Large Language Models: A Survey [73.87651986156006]
  We present a comprehensive survey of bias evaluation and mitigation techniques for large language models (LLMs) We first consolidate, formalize, and expand notions of social bias and fairness in natural language processing. We then unify the literature by proposing three intuitive, two for bias evaluation, and one for mitigation.
  arXiv  Detail & Related papers  (2023-09-02T00:32:55Z)
• Queer People are People First: Deconstructing Sexual Identity Stereotypes in Large Language Models [3.974379576408554]
  Large Language Models (LLMs) are trained primarily on minimally processed web text. LLMs can inadvertently perpetuate stereotypes towards marginalized groups, like the LGBTQIA+ community.
  arXiv  Detail & Related papers  (2023-06-30T19:39:01Z)
• CBBQ: A Chinese Bias Benchmark Dataset Curated with Human-AI Collaboration for Large Language Models [52.25049362267279]
  We present a Chinese Bias Benchmark dataset that consists of over 100K questions jointly constructed by human experts and generative language models. The testing instances in the dataset are automatically derived from 3K+ high-quality templates manually authored with stringent quality control. Extensive experiments demonstrate the effectiveness of the dataset in detecting model bias, with all 10 publicly available Chinese large language models exhibiting strong bias in certain categories.
  arXiv  Detail & Related papers  (2023-06-28T14:14:44Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.