arXiv ·
Researchers at MBZUAI release SlimPajama-DC, an empirical analysis of data combinations for pretraining LLMs using the SlimPajama dataset. The study examines the impact of global vs. local deduplication and the proportions of highly-deduplicated multi-source datasets. Results show that increased data diversity after global deduplication is crucial, with the best configuration outperforming models trained on RedPajama.
arXiv ·
A new method is proposed to reduce the verbosity of LLMs in step-by-step reasoning by retaining moderately easy problems during Reinforcement Learning with Verifiable Rewards (RLVR) training. This approach acts as an implicit length regularizer, preventing the model from excessively increasing output length on harder problems. Experiments using Qwen3-4B-Thinking-2507 show the model achieves baseline accuracy with nearly twice shorter solutions.
arXiv ·
MBZUAI researchers introduce SocialMaze, a new benchmark for evaluating social reasoning capabilities in large language models (LLMs). SocialMaze includes six diverse tasks across social reasoning games, daily-life interactions, and digital community platforms, emphasizing deep reasoning, dynamic interaction, and information uncertainty. Experiments show that LLMs vary in handling dynamic interactions, degrade under uncertainty, but can be improved via fine-tuning on curated reasoning examples.
arXiv ·
This paper presents an experience report on teaching an AI course to business executives in the UAE. The course focuses on enabling students to understand how to incorporate AI into existing business processes, rather than focusing only on theoretical and technical aspects. The paper discusses the course overview, curriculum, teaching methods, and reflections on teaching adult learners in the UAE.
arXiv ·
This paper provides an overview of the UrduFake@FIRE2021 shared task, which focused on fake news detection in the Urdu language. The task involved binary classification of news articles into real or fake categories using a dataset of 1300 training and 300 testing articles across five domains. 34 teams registered, with 18 submitting results and 11 providing technical reports detailing various approaches from BoW to Transformer models, with the best system achieving an F1-macro score of 0.679.
arXiv ·
This study assesses workforce preparedness for AI in the GCC region, using socio-technical systems theory to analyze national AI strategies and initiatives in KSA, UAE, Qatar, Kuwait, Bahrain, and Oman. The research combines TF-IDF analysis, case studies of MBZUAI and SDAIA Academy, and scenario planning to evaluate the balance between technical capacity and social alignment. The study identifies a potential two-track talent system and emphasizes the importance of regulatory convergence for successful AI adoption.
arXiv ·
This paper presents a reinforcement learning framework for optimizing energy pricing in peer-to-peer (P2P) energy systems. The framework aims to maximize the profit of all components in a microgrid, including consumers, prosumers, the service provider, and a community battery. Experimental results on the Pymgrid dataset demonstrate the approach's effectiveness in price optimization, considering the interests of different components and the impact of community battery capacity.