Topics ›
Large language models (LLMs) developed, deployed, and researched across the GCC — including Falcon (TII), Jais (MBZUAI/Inception), AceGPT (KAUST), ALLaM (SDAIA), and Fanar (QCRI).
251–300 articles · Page 6 RSS ↗
This paper introduces DetectLLM-LRR and DetectLLM-NPR, two novel zero-shot methods for detecting machine-generated text using log rank information. Experiments across three datasets and seven language models demonstrate improvements of up to 3.9 AUROC points over state-of-the-art methods. The code and data for both methods are available on Github.
This paper introduces ProgramFC, a fact-checking model that decomposes complex claims into simpler sub-tasks using a library of functions. The model uses LLMs to generate reasoning programs and executes them by delegating sub-tasks, enhancing explainability and data efficiency. Experiments on fact-checking datasets demonstrate ProgramFC's superior performance compared to baseline methods, with publicly available code and data.
This paper introduces a Regulatory Knowledge Graph (RKG) for the Abu Dhabi Global Market (ADGM) regulations, constructed using language models and graph technologies. A portion of the regulations was manually tagged to train BERT-based models, which were then applied to the rest of the corpus. The resulting knowledge graph, stored in Neo4j, and code are open-sourced on GitHub to promote advancements in compliance automation.
The paper introduces ORCA, a new public benchmark for evaluating Arabic language understanding. ORCA covers diverse Arabic varieties and includes 60 datasets across seven NLU task clusters. The benchmark was used to compare 18 multilingual and Arabic language models and includes a public leaderboard with a unified evaluation metric. Why it matters: ORCA addresses the lack of a comprehensive Arabic benchmark, enabling better progress measurement for Arabic and multilingual language models.
The paper introduces LLMEffiChecker, a tool to test the computational efficiency robustness of LLMs by identifying vulnerabilities that can significantly degrade performance. LLMEffiChecker uses both white-box (gradient-guided perturbation) and black-box (causal inference-based perturbation) methods to delay the generation of the end-of-sequence token. Experiments on nine public LLMs demonstrate that LLMEffiChecker can substantially increase response latency and energy consumption with minimal input perturbations.
This paper introduces a unified deep autoregressive model (UAE) for cardinality estimation that learns joint data distributions from both data and query workloads. It uses differentiable progressive sampling with the Gumbel-Softmax trick to incorporate supervised query information into the deep autoregressive model. Experiments show UAE achieves better accuracy and efficiency compared to state-of-the-art methods.
The paper introduces AraELECTRA, a new Arabic language representation model. AraELECTRA is pre-trained using the replaced token detection objective on large Arabic text corpora. The model is evaluated on multiple Arabic NLP tasks, including reading comprehension, sentiment analysis, and named-entity recognition. Why it matters: AraELECTRA outperforms current state-of-the-art Arabic language representation models, given the same pretraining data and even with a smaller model size, advancing Arabic NLP.
The paper introduces AraGPT2, a suite of pre-trained transformer models for Arabic language generation, with the largest model (AraGPT2-mega) containing 1.46 billion parameters. Trained on a large Arabic corpus of internet text and news, AraGPT2-mega demonstrates strong performance in synthetic news generation and zero-shot question answering. To address the risk of misuse, the authors also released a discriminator model with 98% accuracy in detecting AI-generated text. Why it matters: This release of both the model and discriminator fills a critical gap in Arabic NLP and encourages further research and applications in the field.
arXiv · · Healthcare Research
A retrospective study in Abu Dhabi, UAE, developed a machine learning-based prognostic system to predict the risk of seven complications in COVID-19 patients using data from 3,352 patient encounters. The system, trained on data from the first 24 hours of admission, achieved high accuracy (AUROC > 0.80) in predicting complications like AKI, ARDS, and elevated biomarkers in geographically split test sets. The models primarily used gradient boosting and logistic regression.
This paper introduces GigaBERT, a customized bilingual BERT model pre-trained for Arabic NLP and English-to-Arabic zero-shot transfer learning. The study evaluates GigaBERT's performance on four information extraction tasks: named entity recognition, part-of-speech tagging, argument role labeling, and relation extraction. Results show that GigaBERT outperforms mBERT, XLM-RoBERTa, and AraBERT in both supervised and zero-shot transfer settings. Why it matters: GigaBERT advances Arabic NLP by providing a high-performing, publicly available model tailored for the complexities of the Arabic language and cross-lingual applications.
Researchers at the American University of Beirut (AUB) have released AraBERT, a BERT model pre-trained specifically for Arabic language understanding. The model was trained on a large Arabic corpus and compared against multilingual BERT and other state-of-the-art methods. AraBERT achieved state-of-the-art performance on several tested Arabic NLP tasks including sentiment analysis, named entity recognition, and question answering. Why it matters: This release provides the Arabic NLP community with a high-performing, open-source language model, facilitating further research and development.
Researchers introduce AraNet, a deep learning toolkit for Arabic social media processing. The toolkit uses BERT models trained on social media datasets to predict age, dialect, gender, emotion, irony, and sentiment. AraNet achieves state-of-the-art or competitive performance on these tasks without feature engineering. Why it matters: The public release of AraNet accelerates Arabic NLP research by providing a comprehensive, deep learning-based tool for various social media analysis tasks.
This paper introduces a large-scale historical corpus of written Arabic spanning 1400 years. The corpus was cleaned and processed using Arabic NLP tools, including identification of reused text. The study uses a novel automatic periodization algorithm to study the history of the Arabic language, confirming the division into Modern Standard and Classical Arabic. Why it matters: This resource enables further computational research into the evolution of Arabic and the development of NLP tools for historical texts.
G42 · · Infrastructure Partnership
Core42, a G42 company specializing in sovereign cloud and AI infrastructure, announced a strategic partnership with Solutions+, an AI-driven shared services company for the MIC Group, at Make it in the Emirates in Abu Dhabi. Core42 will serve as the foundational infrastructure provider, offering GPU compute and large language model access via its Compass API, while Solutions+ will be the primary implementation and data services partner, leveraging this infrastructure for its WEAVE AI platform. The agreement establishes a framework for delivering secure and scalable AI infrastructure across MIC Group companies and Abu Dhabi government entities, with broader scope for government and private sector clients. Why it matters: This partnership significantly advances the UAE's sovereign AI capabilities and accelerates enterprise AI adoption by combining robust infrastructure with specialized implementation expertise for critical national sectors.
A new paper coauthored by researchers at The University of Melbourne and MBZUAI explores disagreement in human annotation for AI training. The paper treats disagreement as a signal (human label variation or HLV) rather than noise, and proposes new evaluation metrics based on fuzzy set theory. These metrics adapt accuracy and F-score to cases where multiple labels may plausibly apply, aligning model output with the distribution of human judgments. Why it matters: This research addresses a key challenge in NLP by accounting for the inherent ambiguity in human language, potentially leading to more robust and human-aligned AI systems.
MBZUAI · · Healthcare LLM
MBZUAI's BiMediX2, a bilingual healthcare multi-modal model, won Meta's Llama Impact Innovation Award for its potential in solving healthcare accessibility challenges across the Middle East and Africa. Built using Llama 3.1, the model understands medical queries in both English and Arabic, interprets medical images, and is integrated as a chatbot on Telegram with speech functionality. The model was also presented at the AI for Sustainable Development Platform Launch Event and integrated into the UNDP for telemedicine. Why it matters: The model's bilingual capabilities and accessibility on low-cost devices and speech-based interaction have potential to improve healthcare access for marginalized populations in the region.
MBZUAI researchers developed LLMVoX, a system enabling LLMs to produce real-time speech, including Arabic. LLMVoX addresses limitations of existing end-to-end and cascaded pipeline approaches, which suffer from either degraded reasoning or latency. LLMVoX was developed as part of Project OMER, which was recently awarded Regional Research Grant from Meta. Why it matters: This enhances the potential of LLMs to function as more natural, multimodal virtual assistants, especially for Arabic-speaking users in the Middle East.
MBZUAI researchers won second place at the AgentX Competition at UC Berkeley for their benchmark measuring AI agents' reasoning across images, comparisons, and video. The Agent-X dataset includes 828 tasks across six domains, requiring agents to use 14 executable tools without explicit instructions. Agent-X analyzes the agent's full reasoning trajectory, unlike typical evaluations that focus only on final answers. Why it matters: The benchmark exposes limitations in current multimodal AI agents and provides a more rigorous evaluation framework for real-world applications in the region and beyond.
MBZUAI · · Healthcare Research
MBZUAI researchers developed MedAgentSim, a simulated hospital environment to evaluate AI diagnostic abilities. The simulation uses LLM-powered agents to mimic doctor-patient conversations, providing a dynamic assessment of diagnostic skills. The system includes doctor, patient, and evaluator agents that interact within the simulated hospital, making real-time decisions. Why it matters: This research offers a more realistic evaluation of AI in clinical settings, addressing limitations of current benchmarks and potentially improving AI's use in healthcare.
MBZUAI researchers received high honors at EMNLP 2025 for two research papers, placing them in the top 2% of accepted work. One paper, MAviS, is a multimodal AI system that identifies bird species by combining images, sounds, and text. The other award-winning paper focuses on uncertainty in LLM-as-a-Judge. Why it matters: The recognition highlights MBZUAI's growing influence in NLP and multimodal AI research, particularly in domain-specific applications like biodiversity conservation.
MBZUAI researchers have developed MAviS, a new multimodal dataset, benchmark, and chatbot for fine-grained bird species recognition. MAviS includes images, audio, and text to help models identify subtle differences between species, especially rare and regional varieties. The related study was presented at EMNLP 2025 and selected as a "Senior Area Chair Highlight". Why it matters: This work addresses a key limitation in AI's ability to support biodiversity conservation and ecological monitoring in the region and globally.
Google.org is providing $1 million to MBZUAI to fund a research initiative led by Professor Thamar Solorio focused on addressing the “data divide” in AI for underrepresented languages, especially those in the MENA region. The project aims to create resource-lean AI models tailored to the sociocultural and linguistic realities of MENA, requiring less data and computational power. This initiative will also support the training of postdoctoral and early-career researchers at MBZUAI. Why it matters: The funding will help create AI technologies grounded in the linguistic nuances of the MENA region, rather than adapting Western models, while also democratizing AI development by lowering resource requirements.
MBZUAI, G42, and Cerebras Systems have launched K2 Think V2, a 70-billion parameter reasoning system built on the K2-V2 base model. K2 Think V2 is fully open-source, from pre-training data to post-training alignment, ensuring transparency and reproducibility. It achieves leading results on complex reasoning benchmarks like AIME2025 and GPQA-Diamond. Why it matters: This release marks a significant advancement in the UAE's AI capabilities, demonstrating leadership in building globally accessible and fully sovereign AI systems focused on reasoning.
MBZUAI's Institute of Foundation Models (IFM) has released K2 Think V2, a 70 billion parameter open-source general reasoning model built on K2 V2 Instruct. The model excels in complex reasoning benchmarks like AIME2025 and GPQA-Diamond, and features a low hallucination rate with long context reasoning capabilities. K2 Think V2 is fully sovereign and open, from pre-training through post-training, using IFM-curated data and a Guru dataset. Why it matters: This release contributes to closing the gap between community-owned reproducible AI and proprietary models, particularly in reasoning and long-context understanding for Arabic NLP tasks.
MBZUAI · · Healthcare Research
MBZUAI researchers co-led a study published in Nature demonstrating that GluFormer, an AI foundation model trained on continuous glucose monitoring (CGM) data, more accurately predicts long-term diabetes and cardiovascular risk than current clinical standards. GluFormer, built on a transformer architecture and trained using NVIDIA AI infrastructure on over 10 million CGM measurements, forecasts individual health risks using short-term glucose dynamics. In a 12-year follow-up, the model captured 66% of new-onset diabetes cases and 69% of cardiovascular-death events in its highest-risk group, outperforming established CGM-derived metrics across 19 external cohorts. Why it matters: The development of GluFormer represents a significant advancement in personalized healthcare, enabling proactive and individualized health strategies through the analysis of dynamic glucose data.
Researchers from MBZUAI developed "uncertainty quantification heads" (UQ heads) to detect hallucinations in language models by probing internal states and estimating the credibility of generated text. UQ heads leverage attention maps and logits to identify potential hallucinations without altering the model's generation process or relying on external knowledge. The team found that UQ heads achieved state-of-the-art performance in claim-level hallucination detection across different domains and languages. Why it matters: This approach offers a more efficient and accurate method for identifying hallucinations, improving the reliability and trustworthiness of language models in various applications.
MBZUAI faculty member Ekaterina Kochmar and postdoctoral researcher Kaushal Kumar Maurya won a Google Academic Research Award for their research on an intelligent tutoring system. The project, "2σ-ITS," aims to develop an educational foundation model for personalized learning and to support tutors in reaching students with limited access to mainstream education. The Google award provides funding and collaboration opportunities for researchers, with Kochmar and Maurya being the only team from the Middle East to win. Why it matters: This award highlights the growing recognition of AI's potential to improve educational equity and access in the region and beyond.
MBZUAI researchers presented a method for cross-cultural transfer learning to improve language models' understanding of diverse Arab cultures. They used in-context learning and demonstration-based reinforcement (DITTO) to transfer cultural knowledge between countries. Experiments showed up to 34% improvement in performance on cultural understanding benchmarks using only a few demonstrations. Why it matters: This research addresses the gap in cultural understanding of Arabic language models, especially for smaller Arab countries, and provides a novel transfer learning approach.
Researchers from MBZUAI, Carnegie Mellon University, and Meta AI presented a new approach called ThoughtComm at NeurIPS 2025 where AI agents communicate through internal, latent representations instead of natural language. This framework extracts and selectively shares latent "thoughts" from agents' internal states, representing the underlying structure of their reasoning. Results show that agents coordinate more effectively, reach consensus faster, and solve problems more accurately using this method. Why it matters: Bypassing the limitations of natural language in AI communication could lead to more efficient and accurate multi-agent systems, impacting areas like robotics, collaborative AI, and distributed problem-solving.
IFM has released K2-V2, a 70B-class LLM that takes a "360-open" approach by making its weights, data, training details, checkpoints, and fine-tuning recipes publicly available. K2-V2 matches leading open-weight model performance while offering full transparency, contrasting with proprietary and semi-open Chinese models. Independent evaluations show K2 as a high-performance, fully open-source alternative in the AI landscape. Why it matters: K2-V2 provides developers with a transparent and reproducible foundation model, fostering trust and enabling customization without sacrificing performance, which is crucial for sensitive applications in the region.
Inception, Cerebras, and MBZUAI have released Jais 2, a 70 billion parameter open-weight Arabic LLM. Jais 2 is trained on an Arabic-first dataset and features a redesigned architecture for stronger reasoning and fluency across Arabic dialects and English. It integrates a safety-first framework and demonstrates capabilities in understanding Arabic poetry, culture, and social media tone. Why it matters: Jais 2 addresses the historical underrepresentation of Arabic in AI by providing a culturally and linguistically faithful model, potentially accelerating innovation across the region.
MBZUAI's Institute of Foundation Models has released K2, a 70-billion-parameter, reasoning-centric foundation model. K2 is designed to be fully inspectable, with open weights, training code, data composition, mid-training checkpoints, and evaluation harnesses. K2 outperforms Qwen2.5-72B and approaches the performance of Qwen3-235B. Why it matters: This release promotes transparency and reproducibility in AI development, providing researchers with the resources needed to study, adapt, and build upon a strong foundation model.
MBZUAI researchers at the Institute of Foundation Models (IFM) investigated the role of reinforcement learning (RL) in improving reasoning abilities of language models. Their study found that RL acts as an 'elicitor' for reasoning in domains frequently encountered during pre-training (e.g., math, coding), while genuinely teaching new reasoning skills in underrepresented domains (e.g., logic, simulations). To support their analysis, they created a new dataset called GURU containing 92,000 examples across six domains. Why it matters: This research clarifies the impact of reinforcement learning on language model reasoning, paving the way for developing models with more generalizable reasoning abilities across diverse domains, an important direction for more capable AI systems.
MBZUAI researchers created Open CaptchaWorld, a new benchmark to test AI agents on solving CAPTCHAs. The benchmark includes 20 modern CAPTCHA types that require perception, reasoning, and interactive actions within a browser. While humans achieve 93.3% accuracy, the best AI agent only reaches 40% on the benchmark. Why it matters: This research highlights a critical gap in current AI agent capabilities, as CAPTCHAs are gatekeepers to high-value web actions like e-commerce and secure logins.
MBZUAI researchers presented a NeurIPS 2024 Spotlight paper that quantifies AI vulnerability by measuring bits leaked per query. Their formula predicts the minimum queries needed for attacks based on mutual information between model output and attacker's target. Experiments across seven models and three attack types (system-prompt extraction, jailbreaks, relearning) validate the relationship. Why it matters: This work offers a framework to translate UI choices (like exposing log-probs or chain-of-thought) into concrete attack surfaces, informing more secure AI design and deployment in the region.
MBZUAI's Institute of Foundation Models (IFM) has developed PAN, a novel interactive world model capable of long-horizon simulation. PAN uses a Generative Latent Prediction (GLP) architecture, coupling internal latent reasoning with generative supervision in the visual domain. The model evolves an internal latent state conditioned on history and natural language actions, then decodes that state into a video segment using a Causal Swin-DPM mechanism for smooth transitions. Why it matters: PAN represents a significant advancement in AI's ability to simulate and predict evolving environments, enabling more steerable and coherent long-term video generation and opening new possibilities for interactive AI systems.
MBZUAI · · Research Healthcare
MBZUAI researchers have developed MorphDiff, a diffusion model that predicts cell morphology from gene expression data. MorphDiff uses the transcriptome to generate realistic post-perturbation images, either from scratch or by transforming a control image. The model combines a Morphology Variational Autoencoder (MVAE) with a Latent Diffusion Model, enabling both gene-to-image generation and image-to-image transformation. Why it matters: This could significantly accelerate drug discovery and biological research by allowing scientists to preview cellular changes before conducting experiments.
A new paper from MBZUAI demonstrates that state-of-the-art speech models can be easily jailbroken using audio perturbations to generate harmful content, achieving success rates of 76-93% on models like Qwen2-Audio and LLaMA-Omni. The researchers adapted projected gradient descent (PGD) to the audio domain to optimize waveforms that push the model towards harmful responses. They propose a defense mechanism based on post-hoc activation patching that hardens models at inference time without retraining. Why it matters: This research highlights a critical vulnerability in speech-based LLMs and offers a practical solution, contributing to the development of more secure and trustworthy AI systems in the region and globally.
MBZUAI researchers introduced Droid, a resource suite and detector family, at EMNLP 2025 designed to distinguish between AI-generated and human-written code. The project addresses the challenge of identifying AI-generated code in software development, considering the prevalence of AI-suggested code and the risks of obfuscated backdoors and feedback loops. DroidCollection includes over one million code samples across seven programming languages, three coding domains, and outputs from 43 different code models, including human-AI co-authored code and adversarially humanized machine code. Why it matters: This research is crucial for maintaining software security and integrity in the age of AI-assisted coding, providing a robust tool for detecting AI-generated code across diverse languages and domains.
A new paper at ICCV 2025, co-authored by MBZUAI Ph.D. student Dmitry Demidov, introduces Dense-WebVid-CoVR, a 1.6-million sample benchmark for composed video retrieval (CoVR). The benchmark features longer, context-rich descriptions and modification texts, generated using Gemini Pro and GPT-4o, with manual verification. The paper also presents a unified fusion approach that jointly reasons across video and text inputs, improving performance on fine-grained edit details. Why it matters: This work advances video search capabilities by enabling more human-like queries, which is crucial for creative and analytic workflows that require nuanced video retrieval.
MBZUAI · · Research Healthcare
MBZUAI and GenBio AI have won the UAE AI Award 2025 in the AI Scientific Research category for their "Unified Protein Language Modeling Framework". The winning project enables AI to learn protein function, generate sequences, and predict 3D structures. The AI-driven simulation approach aims to accelerate drug development, reduce costs, and improve success rates. Why it matters: This award highlights the UAE's commitment to fostering AI innovation in biomedicine and drug discovery, positioning the region as a leader in AI-driven healthcare advancements.
MBZUAI · · Healthcare Research
MBZUAI researchers Darya Taratynova and Shahad Hardan developed Forget-MI, a method for making clinical AI models "unlearn" specific patient data without retraining the entire model. Forget-MI addresses the challenge of removing patient data from AI models trained on multimodal records (like chest X-rays and reports) due to regulations like GDPR and HIPAA. The method unlearns both unimodal (image or text) and joint (image-text) associations while retaining overall accuracy using a late-fusion multimodal classifier. Why it matters: This research provides a practical solution to a critical privacy concern in healthcare AI, enabling compliance with data protection regulations and fostering trust in AI-driven medical applications.
MBZUAI · · AI Data Analytics
MBZUAI researchers developed Data Wise, an AI platform that provides instant data analytics for businesses. The platform uses AI agents and LLMs to analyze raw customer data and generate actionable recommendations. Data Wise aims to address the shortage of data scientists, particularly in the UAE and GCC. Why it matters: This platform democratizes access to advanced analytics, empowering businesses in the region to make data-driven decisions without relying on scarce technical expertise.
MBZUAI awarded OpenAI CEO Sam Altman an honorary doctorate in recognition of his leadership in AI. The ceremony was attended by UAE dignitaries and university leadership, who emphasized Altman's impact on making AI accessible globally. Altman expressed his honor and commended MBZUAI's commitment to AI research and education. Why it matters: This award recognizes the growing importance of AI and the UAE's ambition to be a leader in the field, highlighting the alignment between global AI innovators and the nation's vision.
MBZUAI and G42 have launched K2 Think, an open-source AI system for advanced reasoning with 32 billion parameters. It outperforms reasoning models 20 times larger, employing techniques like long chain-of-thought fine-tuning and reinforcement learning. K2 Think will be available on Cerebras' platform, achieving 2,000 tokens per second, and ranks highly in math performance. Why it matters: This launch positions the UAE as a leader in AI innovation through public-private partnerships and open-source contributions, demonstrating that efficient AI design can rival larger models.
MBZUAI Ph.D. students Muhammad Maaz and Hanoona Rasheed interned at Meta, developing a vision encoder for images and videos. They created PerceptionLM, a multimodal language model, to generate synthetic video-caption data to train the Perception Encoder. The team addressed the challenge of limited labeled video data by building a multimodal language model called PerceptionLM to understand video's spatial and temporal aspects. Why it matters: This highlights MBZUAI's strength in computer vision and provides students opportunities to contribute to cutting-edge research at global tech firms.
Researchers at MBZUAI have developed a new automatic method to examine cross-lingual abilities in multilingual language models, testing 10 models across 16 languages. They combined beam search with language-model-based simulation, generating 6,000 bilingual question pairs and found significant performance drops compared to English, even in high-resource languages like Chinese. The method introduces perturbations to test the models' ability to transfer knowledge rather than rely on memorization. Why it matters: This research highlights critical gaps in cross-lingual AI, providing a framework for developing more equitable and effective multilingual models, especially for Arabic and other under-represented languages.
Researchers at MBZUAI and other institutions have published a study at ACL 2024 investigating how jailbreak attacks work on LLMs. The study used a dataset of 30,000 prompts and non-linear probing to interpret the effects of jailbreak attacks, finding that existing interpretations were inadequate. The researchers propose a new approach to improve LLM safety against such attacks by identifying the layers in neural networks where the behavior occurs. Why it matters: Understanding and mitigating jailbreak attacks is crucial for ensuring the responsible and secure deployment of LLMs, particularly in the Arabic-speaking world where these models are increasingly being used.
MBZUAI researchers have created ArabCulture, a new benchmark dataset to measure cultural commonsense reasoning capabilities in Arabic language models. The dataset was built by native Arabic speakers from 13 countries and is the largest of its kind. Testing 31 language models, the researchers found that many systems struggle with understanding cultural concepts across the Arab world. Why it matters: The new benchmark addresses a gap in AI, enabling development of culturally-aware AI systems tailored to the nuances of the Arabic-speaking world.
MBZUAI researchers Nils Lukas and Toluwani Samuel Aremu will present a paper at ICML 2025 demonstrating the vulnerability of current watermarking techniques in LLMs. Their research shows that adaptive paraphrasers can evade detection from watermarks with negligible impact on text quality, costing less than $10 of GPU compute. The attack involves fine-tuning a small open-weight model to rewrite sentences until surrogate keys no longer trigger detection. Why it matters: This work highlights critical weaknesses in current AI provenance methods, suggesting the need for more robust watermarking techniques to maintain trust in the authenticity of AI-generated content.