We present DialectalArabicMMLU, a new benchmark for evaluating the performance of large language models (LLMs) across Arabic dialects. While recently developed Arabic and multilingual benchmarks have advanced LLM evaluation for Modern Standard Arabic (MSA), dialectal varieties remain underrepresented despite their prevalence in everyday communication. DialectalArabicMMLU extends the MMLU-Redux framework through manual translation and adaptation of 3K multiple-choice question–answer pairs into five major dialects (Syrian, Egyptian, Emirati, Saudi, and Moroccan), yielding a total of 15K QA pairs across 32 academic and professional domains (22K QA pairs when also including English and MSA). The benchmark enables systematic assessment of LLM reasoning and comprehension beyond MSA, supporting both task-based and linguistic analysis. We evaluate 19 open-weight Arabic and multilingual LLMs (1B–13B parameters) and report substantial performance variation across dialects, revealing persistent gaps in dialectal generalization. DialectalArabicMMLU provides the first unified, human-curated resource for measuring dialectal understanding in Arabic, thus promoting more inclusive evaluation and future model development.

Dialectal Arabic (DA) remains under-resourced compared to Modern Standard Arabic (MSA), particularly for machine reading comprehension (MRC) and question answering (QA). Existing Arabic QA benchmarks primarily focus on formal written MSA or multiple-choice QA, with limited coverage of naturally spoken dialects. Here, we aim to bridge this gap. We introduce EDRAC, the first large-scale benchmark for dialectal Arabic machine reading comprehension (MRC) and generative QA, covering five major dialects: Egyptian, Moroccan, Emirati, Syrian, and Saudi Arabic. EDRAC contains 499 passages derived from naturally occurring spoken interactions and 4,977 corresponding QA pairs generated through a human–LLM collaborative pipeline combining iterative generation, LLM-as-a-judge evaluation, and human verification. We benchmark Arabic-centric and multilingual LLMs on EDRAC using lexical and semantic metrics. Our results reveal substantial gaps between semantic answer quality and dialectal fidelity, highlighting the limitations of existing evaluation metrics for dialectal Arabic generation. EDRAC provides a realistic and challenging MRC benchmark for future research on dialectal Arabic NLP.

Poetry has long been a central art form for Arabic speakers, serving as a powerful medium of expression and cultural identity. While modern Arabic speakers continue to value poetry, existing research on Arabic poetry within Large Language Models (LLMs) has primarily focused on analysis tasks such as interpretation or metadata prediction, e.g., rhyme schemes and titles. In contrast, our work addresses the practical aspect of poetry creation in Arabic by introducing controllable generation capabilities to assist users in writing poetry. Specifically, we present a large-scale, carefully curated instruction-based dataset in Modern Standard Arabic (MSA) and various Arabic dialects. This dataset enables tasks such as writing, revising, and continuing poems based on predefined criteria, including style and rhyme, as well as performing poetry analysis. Our experiments show that fine-tuning LLMs on this dataset yields models that can effectively generate poetry that is aligned with user requirements, based on both automated metrics and human evaluation with native Arabic speakers. The data and the code are available at https://github.com/mbzuai-nlp/instructpoet-ar

Semantic segmentation is a core component of discourse analysis, yet existing models are primarily developed and evaluated on high-resource written text, limiting their effectiveness on low-resource conversational varieties. In particular, dialectal Arabic exhibits informal syntax, code-switching, and weakly marked discourse structure that challenge standard semantic segmentation approaches for text. In this paper, we introduce a new multi-genre benchmark (more than 1000 samples) for semantic segmentation in Arabic, focusing on dialectal discourse. The benchmark covers casual telephone conversations, code-switched podcasts, expressive dialogue, and broadcast news, and was annotated and validated by native Arabic annotators. Using this benchmark, we show that segmentation models performing well on MSA news genres degrade on dialectal conversational texts. We further propose a segmentation model that targets local semantic coherence and robustness to discourse discontinuities, consistently outperforming strong baselines on dialectal non-news genres. The benchmark and approach generalize to other low-resource spoken languages.

Jais 2 is a family of open-weight, Arabic-centric large language models designed for strong Arabic performance and broad cultural coverage. The family includes, to our knowledge, the largest open Arabic-centric LLM trained from scratch, with 70B parameters, alongside an efficient 8B model. A custom Arabic-centric vocabulary supports efficient training and inference, while an optimized architecture and training recipe improve compute efficiency. Jais 2 achieves the highest scores on the Open Arabic LLM Leaderboard v2 (OALL2) and AraGen v1 among the evaluated open models, while remaining competitive on English benchmarks. It also obtains strong results in Arabic poetry, dream interpretation, translation, and summarization. On the IslamicQA domain, Jais 2-70B achieves the second-highest overall average among the evaluated large open models, while attaining the highest scores on PalmX and the held-out in-house IslamicQA benchmark. We release both model sizes as open-weights on Hugging Face under a commercially permissive license. Jais 2-70B is also available as a chat application on the Web, iOS, and Android, reaching inference speeds of up to 2,000 tokens per second on Cerebras hardware.

Arabic LLMs have improved rapidly, but aligning them with country-specific cultural behavioral preferences remains underexplored. Previous work has focused primarily on instruction-tuning for factual knowledge, overlooking nuanced cultural behaviors. We propose , which to our knowledge is the first large-scale pan-Arab preference data set for the alignment of cultural behavior conditioned on the country. Our contribution includes two resources in English and Modern Standard Arabic: a culturally grounded preference dataset for training and evaluation, and a multiple-choice benchmark designed to assess culturally aligned behavior across nations. All test data are validated by native speakers to ensure authenticity. To optimize cultural alignment, we experiment with post-training methods across multilingual and Arabic-centric language models, evaluating performance on both generation tasks and our cultural multiple-choice benchmark. By releasing both a training dataset and an evaluation benchmark, we provide a foundation for systematically studying and improving culturally aligned behavior in language models.

There is a significant gap in evaluating cultural reasoning in LLMs using conversational datasets that capture culturally rich and dialectal contexts. Most Arabic benchmarks focus on short text snippets in Modern Standard Arabic (MSA), overlooking the cultural nuances that naturally arise in dialogues. To address this gap, we introduce ArabCulture-Dialogue, a culturally grounded conversational dataset covering 13 Arabic-speaking countries, in both MSA and each country’s respective dialect, spanning 12 daily-life topics and 54 fine-grained subtopics. We utilize the dataset to form three benchmarking tasks: (i) multiple-choice cultural reasoning, (ii) machine translation between MSA and dialects, and (iii) dialect-steering generation. Our experiments indicate that the performance gap between MSA and Arabic dialects still exists, whereby the models perform worse on all three tasks in the dialectal setup, compared to the MSA one.

The development of large-scale Visual Question Answering (VQA) datasets has traditionally relied on resource-intensive manual annotation. In addition, most of the existing Arabic VQA datasets focus on culturally-specific and dialect-aware domains. To address these limitations, we propose a new pipeline that leverages Wikipedia template tags to extract the relevant information for each image, which is subsequently utilized by the Large Language Model (LLM) to synthetically generate a new visual question answering dataset. Using this pipeline, we have constructed AraVQA, the most comprehensive Arabic Factoid Visual Question Answering dataset, containing more than 50,000 questions and covering over 20 varied primary subjects within Arabic general knowledge. Our detailed analysis shows that our dataset can serve as a post-training dataset to enhance the performance of existing Visual Language Models (VLMs) on Arabic VQA tasks. Furthermore, we present a novel benchmark, derived from our dataset and validated through manual annotation, that poses more challenges to Arabic VLMs than existing Arabic VQA datasets.

High-quality WordNets are crucial for achieving high-quality results in NLP applications that rely on such resources. However, the wordnets of most languages suffer from serious issues of correctness and completeness with respect to the words and word meanings they define, such as incorrect lemmas, missing glosses and example sentences, or an inadequate, Western-centric representation of the morphology and the semantics of the language. Previous efforts have largely focused on increasing lexical coverage while ignoring other qualitative aspects. In this paper, we focus on the Arabic language and introduce a major revision of the Arabic WordNet that addresses multiple dimensions of lexico-semantic resource quality. As a result, we updated more than 58% of the synsets of the existing Arabic WordNet by adding missing information and correcting errors. In order to address issues of language diversity and untranslatability, we also extended the wordnet structure by new elements: phrasets and lexical gaps.

We present an overview of the ArAIEval shared task, organized as part of the first ArabicNLP 2023 conference co-located with EMNLP 2023. ArAIEval offers two tasks over Arabic text: (1) persuasion technique detection, focusing on identifying persuasion techniques in tweets and news articles, and (2) disinformation detection in binary and multiclass setups over tweets. A total of 20 teams participated in the final evaluation phase, with 14 and 16 teams participating in Task 1 and Task 2, respectively. Across both tasks, we observe that fine-tuning transformer models such as AraBERT is the core of majority of participating systems. We provide a description of the task setup, including description of datasets construction and the evaluation setup. We also provide a brief overview of the participating systems. All datasets and evaluation scripts from the shared task are released to the research community. We hope this will enable further research on such important tasks within the Arabic NLP community.

This paper describes a method to enrich lexical resources with content relating to linguistic diversity, based on knowledge from the field of lexical typology. We capture the phenomenon of diversity through the notion of lexical gap and use a systematic method to infer gaps semi-automatically on a large scale, which we demonstrate on the kinship domain. The resulting free diversity-aware terminological resource consists of 198 concepts, 1,911 words, and 37,370 gaps in 699 languages. We see great potential in the use of resources such as ours for the improvement of a variety of cross-lingual NLP tasks, which we illustrate through an application in the evaluation of machine translation systems.

The emergence of Multi-task learning (MTL)models in recent years has helped push thestate of the art in Natural Language Un-derstanding (NLU). We strongly believe thatmany NLU problems in Arabic are especiallypoised to reap the benefits of such models. Tothis end we propose the Arabic Language Un-derstanding Evaluation Benchmark (ALUE),based on 8 carefully selected and previouslypublished tasks. For five of these, we providenew privately held evaluation datasets to en-sure the fairness and validity of our benchmark. We also provide a diagnostic dataset to helpresearchers probe the inner workings of theirmodels.Our initial experiments show thatMTL models outperform their singly trainedcounterparts on most tasks. But in order to en-tice participation from the wider community,we stick to publishing singly trained baselinesonly. Nonetheless, our analysis reveals thatthere is plenty of room for improvement inArabic NLU. We hope that ALUE will playa part in helping our community realize someof these improvements. Interested researchersare invited to submit their results to our online,and publicly accessible leaderboard.

We present a new wordnet resource for Scottish Gaelic, a Celtic minority language spoken by about 60,000 speakers, most of whom live in Northwestern Scotland. The wordnet contains over 15 thousand word senses and was constructed by merging ten thousand new, high-quality translations, provided and validated by language experts, with an existing wordnet derived from Wiktionary. This new, considerably extended wordnet—currently among the 30 largest in the world—targets multiple communities: language speakers and learners; linguists; computer scientists solving problems related to natural language processing. By publishing it as a freely downloadable resource, we hope to contribute to the long-term preservation of Scottish Gaelic as a living language, both offline and on the Web.

This paper describes the solution that we propose on MADAR 2019 Arabic Fine-Grained Dialect Identification task. The proposed solution utilized a set of classifiers that we trained on character and word features. These classifiers are: Support Vector Machines (SVM), Bernoulli Naive Bayes (BNB), Multinomial Naive Bayes (MNB), Logistic Regression (LR), Stochastic Gradient Descent (SGD), Passive Aggressive(PA) and Perceptron (PC). The system achieved competitive results, with a performance of 62.87 % and 62.12 % for both development and test sets.

In this paper we discuss several models we used to classify 25 city-level Arabic dialects in addition to Modern Standard Arabic (MSA) as part of MADAR shared task (sub-task 1). We propose an ensemble model of a group of experimentally designed best performing classifiers on a various set of features. Our system achieves an accuracy of 69.3% macro F1-score with an improvement of 1.4% accuracy from the baseline model on the DEV dataset. Our best run submitted model ranked as third out of 19 participating teams on the TEST dataset with only 0.12% macro F1-score behind the top ranked system.

In this paper we present the Tren-toTeam system which participated to thetask 3 at SemEval-2017 (Nakov et al.,2017).We concentrated our work onapplying Grice Maxims(used in manystate-of-the-art Machine learning applica-tions(Vogel et al., 2013; Kheirabadiand Aghagolzadeh, 2012; Dale and Re-iter, 1995; Franke, 2011)) to ranking an-swers of a question by answers relevancy. Particularly, we created a ranker systembased on relevancy scores, assigned by 3main components: Named entity recogni-tion, similarity score, sentiment analysis. Our system obtained a comparable resultsto Machine learning systems.

WordNet represents polysemous terms by capturing the different meanings of these terms at the lexical level, but without giving emphasis on the polysemy types such terms belong to. The state of the art polysemy approaches identify several polysemy types in WordNet but they do not explain how to classify and organize them. In this paper, we present a novel approach for classifying the polysemy types which exploits taxonomic principles which in turn, allow us to discover a set of polysemy structural patterns.