Researchers introduce AceGPT, a localized large language model (LLM) specifically for Arabic, addressing cultural sensitivity and local values not well-represented in mainstream models. AceGPT incorporates further pre-training with Arabic texts, supervised fine-tuning using native Arabic instructions and GPT-4 responses, and reinforcement learning with AI feedback using a reward model attuned to local culture. Evaluations demonstrate that AceGPT achieves state-of-the-art performance among open Arabic LLMs across several benchmarks. Why it matters: This work advances culturally-aware AI development for Arabic-speaking communities, providing a valuable resource and benchmark for future research.
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.
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 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 presented a study at ACL 2024 on improving Arabic ASR by pre-training on dialectal Arabic. They trained three versions of the ArTST model: one on MSA, one on MSA and dialectal data, and one on MSA, dialectal, and multilingual data. Results showed that pre-training on dialectal Arabic improves ASR performance across MSA and various dialects. Why it matters: This research addresses a key challenge in Arabic NLP, given the diversity and lack of standardization in dialects, which could lead to more accurate speech recognition systems.
Jan Buchmann from TU Darmstadt presented research on NLP for long, structured documents at MBZUAI. The research addresses gaps in using document structure and improving the verifiability of LM responses. Experiments showed that models learn to represent document structure during pre-training, and larger models can cite sources well. Why it matters: This research contributes to making NLP more effective for complex documents like scientific articles and legal texts, which is crucial for information accessibility.
A CMU professor and MBZUAI affiliated faculty presented research on how LLMs store and use knowledge learned during pre-training. The study used a synthetic biography dataset to show that LLMs may not effectively use memorized knowledge at inference time, even with zero training loss. Data augmentation during pre-training can force the model to store knowledge in specific token embeddings. Why it matters: The research highlights limitations in LLM knowledge manipulation and extraction, with implications for improving model architectures and training strategies for more effective knowledge utilization in Arabic LLMs.