This paper introduces an interpretable pipeline that integrates mobility and social media data to analyze human behavior during crises. The framework was evaluated through two case studies, including a longitudinal analysis of UAE COVID-19 behavior from March 2020 to December 2021. The pipeline aligns heterogeneous daily signals, transforms them into binary behavioral states, applies Formal Concept Analysis (FCA) to extract co-occurrence structures, and mines association rules. Results demonstrate clear cross-domain behavioral structures in crises, yielding both scientifically credible and policy-actionable intelligence. Why it matters: This work provides a novel methodological approach for developing actionable crisis management strategies by fusing multimodal data, directly applicable to public health and emergency response in the UAE and the broader region.
Researchers from MBZUAI introduce Forget-MI, a machine unlearning method tailored for multimodal medical data, enhancing privacy by removing specific patient data from AI models. Forget-MI utilizes loss functions and perturbation techniques to unlearn both unimodal and joint data representations. The method demonstrates superior performance in reducing Membership Inference Attacks and improving data removal compared to existing techniques, while preserving overall model performance and enabling data forgetting.
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.
MBZUAI researchers are developing an AI-driven energy management system that optimizes ice battery technology for cooling in hot climates. The system stores energy as frozen water during times of energy surplus and uses it to cool buildings when demand peaks. The AI model integrates multimodal data from weather forecasts, environmental sensors, and power grid signals to determine when to store or release thermal energy. Why it matters: This approach promises to reduce fossil fuel dependence and lower energy costs while improving cooling performance in regions like the UAE.
MBZUAI researchers presented a new causal representation learning framework at ICLR for identifying latent causal variables in multimodal biological data. The framework addresses the challenge of uncovering underlying causal factors from lab tests, genetic information, and medical images. The new approach can identify latent causal variables and their influence on observed biological outcomes across modalities. Why it matters: The model's ability to analyze causal mechanisms between modalities can lead to more complete insights in biomedical research.
Researchers at MBZUAI have developed a new machine learning method called survival rank-n-contrast (SurvRNC) to improve survival models for cancer prognoses. The method is designed to predict survival times for head and neck cancer patients using multimodal data while accounting for censored data (missing values). Numan Saeed presented the team’s work at the 27th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI). Why it matters: Accurate prognoses can significantly improve patient outcomes, and this research contributes to advancements in machine learning techniques for handling complex and incomplete medical data.