The paper "Prompt Revision as a Source of Cultural Bias in Text-to-Image Systems" investigates how commercial text‑to‑image models silently modify user prompts before generating images, a step that is often hidden from users. Using the multilingual benchmark WORLDVIEW, the authors audit the revision layer in DALL‑E‑3, Imagen‑4, and GPT‑Image‑1.5, finding that non‑Western and non‑Anglophone contexts are disproportionately marked, reduced to narrow vocabularies, and stereotyped. The study demonstrates that the revision layer itself is a previously undocumented causal source of cultural stereotyping, underscoring the need to audit deployed systems rather than just the underlying models.
By Aleksandra Urman, Elsa Lichtenegger, Salima Jaoua, Azza Bouleimen, Robin Forsberg, Corinna Hertweck, Stefania Ionescu, Nicol\`o Pagan, Ancsa Hannak, Joachim Baumann
arXiv:2604. 09945v2 Announce Type: replace-cross Abstract: The rapid adoption of large vision-language models (LVLMs) in recent years has been accompanied by growing fairness concerns due to their propensity to reinforce harmful societal stereotypes.
By Phillip Howard, Xin Su, Kathleen C. Fraser
arXiv:2608. 11002v1 Announce Type: cross Abstract: Text-to-image (T2I) generation has achieved remarkable progress in recent years.
By Sicheng Zhang, Zhonghao Yan, Binzhu Xie, Shi Qiu, Muzammal Naseer, Naveed Akhtar, Mubarak Shah
CultureVidBench is a new benchmark that evaluates how well text‑to‑video generation models capture cultural details. It contains 1,000 prompts spanning 12 countries, 6 continents, 8 cultural regions, and 14 cultural aspects, grouped into material culture, social practice & performance, and ritual & ceremony. Human studies and automated assessments show that while current models perform well on semantic adherence and visual quality, they often miss fine‑grained cultural details, especially for underrepresented regions and multimodal cues.
By Xianjing Han, Yuhan Su, Yang Deng, Dong Ma, Wee Peng Tay, Bin Zhu
arXiv:2605. 16716v5 Announce Type: replace-cross Abstract: Text-to-video (T2V) generation has rapidly progressed in visual fidelity, yet its ability to faithfully represent multiple cultures within a single prompt remains underexplored.
By Shuowei Li, Yuming Zhao, Parth Bhalerao, Oana Ignat
The paper "Playing Bias: Epistemic Games for Interpreting Text-to-Image Models" proposes a three-stage, humanities-informed framework—Interpret, Interrogate, and Expose—to make the cultural work of text-to-image (T2I) systems visible and debatable. These epistemic games, tested with undergraduate fine‑arts students, shift discussions from impressionistic reactions to concise, evidence‑based hypotheses, using lightweight classroom tools for reproducibility and structured collaboration. The authors argue that such game‑based approaches foster critical literacy, enabling learners to recognize, discuss, and contest how T2I models encode bias.
By Amalia Foka
arXiv:2601. 22396v2 Announce Type: replace-cross Abstract: Despite the growing utility of Large Language Models (LLMs) for simulating human behavior, the extent to which these synthetic personas accurately reflect world and moral value systems across different cultural conditionings remains uncertain.
By Candida M. Greco, Lucio La Cava, Andrea Tagarelli
Claims about the universality of human concepts have been predominantly assessed through linguistic similarity across languages and cultures. However, words are effective as communication devices because they compress rich experiential variation into shared conventions, potentially obscuring hidden individual and cultural differences in how concepts are mentally represented.
arXiv:2605. 16716v4 Announce Type: replace-cross Abstract: Text-to-video (T2V) generation has rapidly progressed in visual fidelity, yet its ability to faithfully represent multiple cultures within a single prompt remains underexplored.
By Shuowei Li, Yuming Zhao, Parth Bhalerao, Oana Ignat
The study examines how large language models (LLMs) respond to simultaneous cultural identity cues, comparing explicit country names with implicit language cues. Using World Values Survey probing and open‑ended sociopolitical generation, it finds that language cues shift responses more than explicit cues but not toward the named country’s distribution, while explicit cues produce more differentiated value changes. Across 1.6 million responses, naming a country narrows the gap to the model’s own country‑specific answers by 30–44%.
By Siddhesh Pawar, Arnav Arora, Dustin Wright, Dirk Hovy, Isabelle Augenstein
arXiv:2606. 07311v1 Announce Type: cross Abstract: As video generation models like Veo 3.
By Anku Rani, Wei Dai, Shravan Nayak, Pattie Maes, Mahdi M. Kalayeh, Paul Pu Liang
The paper examines how multimodal language models, specifically Qwen3‑VL‑8B‑Instruct, encode different observer perspectives in their internal representations. By varying persona content and prompt framing across 100 images and 53 conditions, the study finds that early layers encode image identity while later layers develop persona‑specific structures. Prompt framing more strongly alters internal representations than persona identity, yet persona content has a larger effect on behavioural judgments, revealing a dissociation between representational geometry and functional outcomes.
By Mathis Immertreu, Fitim Abdullahu, Thomas Kinfe, Claus Metzner, Helmut Grabner, Patrick Krauss, Achim Schilling