analiza aiqbench
This research addresses the critical gap in multimodal models regarding cultural biases by introducing ArtECulture, a new benchmark featuring 6,792 artworks across three cultures to evaluate emotion perception accuracy. The authors demonstrate that current MLLMs struggle with this task without explicit knowledge injection and propose a retrieval-augmented framework using a concept-based cultural knowledge base to improve zero-shot performance. This work is essential for advancing the robustness of AI systems in handling diverse global contexts beyond Western-centric training data.
ArtECulture: Benchmarking Culture-Conditioned Visual Emotion Understanding in Multimodal Large Language Models
arXiv:2608.03358v1 Announce Type: new Abstract: Existing visual emotion understanding methods typically ignore cultural variations in emotional perception. We introduce culture-conditioned visual emotion understanding, a task that predicts the culture-specific emotional perception of a given image and explains the underlying rationale. Although related benchmarks exist, they are limited by inconsistent individual annotations, which hinder the derivation of majority-supported culture-level emotion labels, and imbalanced cultural coverage. Thus, we present ArtECulture, a benchmark containing 6,792 artworks with culture-specific emotion labels and explanations across English, Chinese, and Arabic cultures, with balanced Western and non-Western content. Evaluations of 16 open- and closed-source Multimodal Large Language Models (MLLMs) under a zero-shot setting reveal that the task remains challenging, with the best model achieving below 50\% accuracy. To address this limitation, we introduce a retrieval-augmented culture-conditioned emotion understanding framework, which leverages a concept-based cultural emotion knowledge base to inject explicit cultural knowledge into MLLMs without additional training. The framework improves both culturally aligned emotion prediction and grounded explanation generation. Our benchmark and code will be publicly released.
Abstract:Existing visual emotion understanding methods typically ignore cultural variations in emotional perception. We introduce culture-conditioned visual emotion understanding, a task that predicts the culture-specific emotional perception of a given image and explains the underlying rationale. Although related benchmarks exist, they are limited by inconsistent individual annotations, which hinder the derivation of majority-supported culture-level emotion labels, and imbalanced cultural coverage. Thus, we present ArtECulture, a benchmark containing 6,792 artworks with culture-specific emotion labels and explanations across English, Chinese, and Arabic cultures, with balanced Western and non-Western content. Evaluations of 16 open- and closed-source Multimodal Large Language Models (MLLMs) under a zero-shot setting reveal that the task remains challenging, with the best model achieving below 50\% accuracy. To address this limitation, we introduce a retrieval-augmented culture-conditioned emotion understanding framework, which leverages a concept-based cultural emotion knowledge base to inject explicit cultural knowledge into MLLMs without additional training. The framework improves both culturally aligned emotion prediction and grounded explanation generation. Our benchmark and code will be publicly released.
Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2608.03358 [cs.CL]
(or arXiv:2608.03358v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2608.03358 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 266 cuvinte
Subiecte conexe
The J-Space Debate, Agent Swarms, and Pacing Frontier AI - Digital Minds Newsletter #4
LessWrong · 18 sept. 2026
[AINews] not much happened today
Latent Space · 10 sept. 2026
PalmLeaf-VQA: A Multi-Script Visual Question Answering Benchmark for Historical Palm-Leaf Manuscript Understanding Across Diverse Regions
arXiv · 30 sept. 2026
AVMeme Exam: A Multimodal Multilingual Multicultural Benchmark for LLMs' Contextual and Cultural Knowledge and Thinking
arXiv · 7 oct. 2026