A study investigated the culturally aware risks of Generative AI for youth aged 7-17 in Saudi Arabia, focusing on privacy and safety challenges. Researchers analyzed 736 Reddit posts, 1,262 X (Twitter) posts, and conducted interviews with 31 Saudi participants including youth, parents, and teachers. Findings highlighted context-dependent risks, particularly regarding the disclosure of personal and family information that conflicts with culturally rooted expectations of modesty, privacy, and honor. The study proposes design implications for inclusive, context-sensitive parental controls that align with local cultural norms and values. Why it matters: This research is crucial for developing AI tools and policies that are culturally appropriate and safeguard youth in non-Western contexts like the Middle East.
The paper introduces FanarGuard, a bilingual moderation filter for Arabic and English language models that considers both safety and cultural alignment. A dataset of 468K prompt-response pairs was created and scored by LLM judges on harmlessness and cultural awareness to train the filter. The first benchmark targeting Arabic cultural contexts was developed to evaluate cultural alignment. Why it matters: FanarGuard advances context-sensitive AI safeguards by integrating cultural awareness into content moderation, addressing a critical gap in current alignment techniques.
This survey paper analyzes over 40 benchmarks used to evaluate Arabic large language models, categorizing them into Knowledge, NLP Tasks, Culture and Dialects, and Target-Specific evaluations. It identifies progress in benchmark diversity but also highlights gaps like limited temporal evaluation and cultural misalignment. The paper also examines methods for creating benchmarks, including native collection, translation, and synthetic generation. Why it matters: The survey provides a comprehensive reference for Arabic NLP research and offers recommendations for future benchmark development to better align with cultural contexts.
MBZUAI researchers release 'Fann or Flop', a new benchmark for evaluating Arabic poetry understanding in LLMs. The benchmark covers 12 historical eras and 14 poetic genres, assessing semantic understanding, metaphor interpretation, and cultural context. Evaluation of state-of-the-art LLMs reveals challenges in poetic understanding despite strong performance on standard Arabic benchmarks.
Researchers introduce AraDiCE, a benchmark for Arabic Dialect and Cultural Evaluation, comprising seven synthetic datasets in various dialects and Modern Standard Arabic (MSA). The benchmark includes approximately 45,000 post-edited samples and evaluates LLMs on dialect comprehension, generation, and cultural awareness across the Gulf, Egypt, and Levant. Results show that Arabic-specific models like Jais and AceGPT outperform multilingual models on dialectal tasks, but challenges remain in dialect identification, generation, and translation. Why it matters: This benchmark and associated datasets will help improve LLMs' ability to understand and generate diverse Arabic dialects and cultural contexts, addressing a significant gap in current models.
A new paper from MBZUAI introduces JEEM, a benchmark dataset for evaluating vision-language models on their understanding of images grounded in four Arabic-speaking societies (Jordan, UAE, Egypt, and Morocco) and their ability to use local dialects. The dataset comprises 2,178 images and 10,890 question-answer pairs reflecting everyday life and culturally specific scenes. Evaluation of several Arabic-capable models (Maya, PALO, Peacock, AIN, AyaV) and GPT-4o revealed that while models can generate fluent language, they struggle with genuine understanding, consistency, and relevance, especially when cultural context is important. Why it matters: This research highlights the challenges of building AI systems that can truly understand and interact with diverse cultures, emphasizing the need for culturally grounded datasets and evaluation metrics.
MBZUAI is conducting research to improve cross-cultural understanding using AI, including studying LLM limitations in recognizing cultural references. They developed "Culturally Yours," a tool that helps users comprehend cultural references in text, and the "All Languages Matter Benchmark" (ALM Bench) to evaluate multimodal LLMs across 100 languages. MBZUAI has also developed LLMs tailored to low-resource languages like Jais (Arabic), Nanda (Hindi), and Sherkala (Kazakh). Why it matters: These initiatives promote inclusivity and ensure AI systems are culturally aware and can serve diverse populations effectively, particularly in the Middle East's multicultural context.
MBZUAI researchers have released ALM Bench, a new benchmark dataset for evaluating the performance of multimodal LLMs on cultural visual question-answer tasks across 100 languages. The dataset includes over 22,000 question-answer pairs across 19 categories, with a focus on low-resource languages and cultural nuances, including three Arabic dialects. They tested 16 open- and closed-source multimodal LLMs on it, revealing a significant need for greater cultural and linguistic inclusivity. Why it matters: The benchmark aims to improve the inclusivity of multimodal AI systems by addressing the underrepresentation of low-resource languages and cultural contexts.