Researchers from MBZUAI have introduced WR-Arena, a new comprehensive benchmark designed to evaluate World Models (WMs) beyond traditional next-state prediction and visual fidelity. WR-Arena assesses WMs across three core dimensions: Action Simulation Fidelity, Long-horizon Forecast, and Simulative Reasoning and Planning, using a curated task taxonomy and diverse datasets. Extensive experiments with state-of-the-art WMs revealed a significant gap between current models' capabilities and human-level hypothetical reasoning. Why it matters: This benchmark provides a critical diagnostic tool and guideline for developing more robust and intelligent world models capable of advanced understanding, forecasting, and purposeful action, particularly for AI research in the region.
A new approach to composed video retrieval (CoVR) is presented, which leverages large multimodal models to infer causal and temporal consequences implied by an edit. The method aligns reasoned queries to candidate videos without task-specific finetuning. A new benchmark, CoVR-Reason, is introduced to evaluate reasoning in CoVR.
The Abu Dhabi Autonomous Racing League (A2RL) Season 2 Grand Final took place at Yas Marina Circuit, featuring six fully driverless racecars. Germany’s TUM team won the championship, followed by TII Racing (UAE) and PoliMOVE (Italy). The event included a Human vs AI showdown and showcased speeds over 250 km/h and advanced AI decision-making. Why it matters: A2RL demonstrates the UAE's commitment to advancing autonomous systems and fostering public trust in AI technologies for various sectors.
Technology Innovation Institute (TII) has developed AI-powered autonomous drones capable of navigating complex environments at speeds up to 80 km/h using only a camera and IMU sensor. The drones use onboard AI-driven visual odometry and reinforcement learning to adapt to their environment in real time. In direct competition, the TII drone set a best lap time of 4.38s, compared to 6.32s and 5.34s for human pilots. Why it matters: This research demonstrates the potential of AI-powered UAVs to surpass human-operated drones in agility and precision, with applications for the transport of goods and potentially people.
TII's Autonomous Robotics Research Center (ARRC) and NYU's Agile Robotics and Perception Lab have released RLtools, an open-source reinforcement learning library. RLtools achieves a 75x speed-up in training compared to existing libraries, enabling drone controller training on standard computers. It allows training on consumer-grade laptops or directly on microcontrollers, addressing resource efficiency and deployment challenges. Why it matters: This library accelerates the development and deployment of autonomous systems by reducing training time and resource requirements, making advanced AI more accessible.
ARRC researchers in collaboration with the University of Bologna and ETH Zürich have developed a CNN-based AI deck to enable autonomous navigation of a 27g nano-drone in unknown environments. The CNN allows the drone to recognize and avoid obstacles using only an onboard camera, running 10x faster and using 10x less memory than previous versions. The demo also featured a swarm of nano-drones flying in formation using ultra-wideband communication. Why it matters: This advancement could significantly enhance the capabilities of nano-drones for applications such as disaster response, where quick and efficient intervention is crucial.
The Technology Innovation Institute (TII) in Abu Dhabi has launched a cloud API providing access to quantum-inspired algorithms developed by its Quantum Research Center (QRC). The platform offers a testbed for partners to evaluate and build proof-of-concept applications, with the first algorithm being a quantum annealing emulator. Access is provided through two interfaces, enabling large-scale classical simulations and supporting the solution of combinatorial optimization problems. Why it matters: This initiative expands TII's quantum ecosystem and facilitates applied research and early-stage industry experimentation with advanced computational methods in the GCC region.
The Technology Innovation Institute (TII) in Abu Dhabi, in collaboration with NVIDIA, has demonstrated large-scale simulations of the adiabatic quantum annealing (QA) algorithm for problem instances involving up to 500,000 qubits. TII's simulator achieved solution quality exceeding that of all solvers evaluated from the MQLib repository, a library for combinatorial optimization benchmarking. The emulator is accessible to external users via an experimental cloud platform hosted at https://q-inspired.tii.ae. Why it matters: This collaboration expands the range of complex optimization problems that can be investigated using quantum-inspired approaches, beyond those currently achievable with near-term quantum hardware.
This study introduces a reinforcement learning (RL) framework using Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) to optimize the cleaning schedules of photovoltaic panels in arid regions. Applied to a case study in Abu Dhabi, the PPO-based framework demonstrated up to 13% cost savings compared to simulation optimization methods by dynamically adjusting cleaning intervals based on environmental conditions. The research highlights the potential of RL in enhancing the efficiency and reducing the operational costs of solar power generation.
This paper introduces ADR-VINS, a monocular visual-inertial state estimation framework based on an Error-State Kalman Filter (ESKF) designed for autonomous drone racing, integrating direct pixel reprojection errors from gate corners as innovation terms. It also introduces ADR-FGO, an offline Factor-Graph Optimization framework for generating high-fidelity reference trajectories for post-flight evaluation in GNSS-denied environments. Validated on the TII-RATM dataset, ADR-VINS achieved an average RMS translation error of 0.134 m and was successfully deployed in the A2RL Drone Championship Season 2. Why it matters: The framework provides a robust and efficient solution for drone state estimation in challenging racing environments, and enables performance evaluation without relying on external localization systems.
The paper introduces ILION, a deterministic execution gate designed to ensure the safety of autonomous AI agents by classifying proposed actions as either BLOCK or ALLOW. ILION uses a five-component cascade architecture that operates without statistical training, API dependencies, or labeled data. Evaluation against existing text-safety infrastructures demonstrates ILION's superior performance in preventing unauthorized actions, achieving an F1 score of 0.8515 with sub-millisecond latency.
The paper presents MonoRace, an onboard drone racing approach using a monocular camera and IMU. The system combines neural-network-based gate segmentation with a drone model for robust state estimation, along with offline optimization using gate geometry. MonoRace won the 2025 Abu Dhabi Autonomous Drone Racing Competition (A2RL), outperforming AI teams and human world champions, reaching speeds up to 100 km/h. Why it matters: This demonstrates a significant advancement in autonomous drone racing, achieving champion-level performance with a resource-efficient monocular system, validated in a real-world competition setting in the UAE.
The paper introduces Yet another Policy Optimization (YaPO), a reference-free method for learning sparse steering vectors in the latent space of a Sparse Autoencoder (SAE) to steer LLMs. By optimizing sparse codes, YaPO produces disentangled, interpretable, and efficient steering directions. Experiments show YaPO converges faster, achieves stronger performance, exhibits improved training stability and preserves general knowledge compared to dense steering baselines.
This paper details the autonomous drone racing system developed for the Abu Dhabi Autonomous Racing League (A2RL) x Drone Champions League competition. The system uses drift-corrected monocular Visual-Inertial Odometry (VIO) fused with YOLO-based gate detection for global position measurements, managed via Kalman filter. A perception-aware planner generates trajectories balancing speed and gate visibility. Why it matters: The system's podium finishes validate the effectiveness of monocular vision-based autonomous drone flight and showcases advancements in AI-powered robotics within the UAE.
The paper introduces OmniGen, a unified framework for generating aligned multimodal sensor data for autonomous driving using a shared Bird's Eye View (BEV) space. It uses a novel generalizable multimodal reconstruction method (UAE) to jointly decode LiDAR and multi-view camera data through volume rendering. The framework incorporates a Diffusion Transformer (DiT) with a ControlNet branch to enable controllable multimodal sensor generation, demonstrating good performance and multimodal consistency.
Researchers at MBZUAI introduce "Interactive Video Reasoning," a new paradigm enabling models to actively "think with videos" by performing iterative visual actions to gather and refine evidence. They developed Video CoM, which reasons through a Chain of Manipulations (CoM), and constructed Video CoM Instruct, an 18K instruction tuning dataset for multi-step manipulation reasoning. The model is further optimized via reinforcement learning with reasoning aware Group Relative Policy Optimization (GRPO), achieving strong results across nine video reasoning benchmarks.
Researchers at MBZUAI have introduced Video-R2, a reinforcement learning approach to improve the consistency and visual grounding of reasoning in multimodal language models. Video-R2 combines timestamp-aware supervised fine-tuning with Group Relative Policy Optimization (GRPO) guided by a Temporal Alignment Reward (TAR). The model demonstrates higher Think Answer Consistency (TAC), Video Attention Score (VAS), and accuracy across multiple benchmarks, showing improved temporal alignment and reasoning coherence for video understanding.
Researchers at MBZUAI have introduced EvoLMM, a self-evolving framework for large multimodal models that enhances reasoning capabilities without human-annotated data or reward distillation. EvoLMM uses two cooperative agents, a Proposer and a Solver, which generate image-grounded questions and solve them through internal consistency, using a continuous self-rewarding process. Evaluations using Qwen2.5-VL as the base model showed performance gains of up to 3% on multimodal math-reasoning benchmarks like ChartQA, MathVista, and MathVision using only raw training images.
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.
Researchers introduce MATRIX, a vision-centric agent tuning framework for robust tool-use reasoning in VLMs. The framework includes M-TRACE, a dataset of 28.5K multimodal tasks with 177K verified trajectories, and Pref-X, a set of 11K automatically generated preference pairs. Experiments show MATRIX consistently outperforms open- and closed-source VLMs across three benchmarks.
Researchers propose MS-NN-steer, a model-structured neural network for autonomous vehicle steering control that integrates nonlinear vehicle dynamics. The controller was validated using real-world data from the Abu Dhabi Autonomous Racing League (A2RL) competition. MS-NN-steer demonstrates improved accuracy, generalization, and robustness compared to general-purpose NNs and the A2RL winning team's controller. Why it matters: This research demonstrates a promising approach to developing transparent and reliable AI for safety-critical autonomous racing applications in the UAE.
This paper introduces Adaptive Entropy-aware Optimization (AEO), a new framework to tackle Multimodal Open-set Test-time Adaptation (MM-OSTTA). AEO uses Unknown-aware Adaptive Entropy Optimization (UAE) and Adaptive Modality Prediction Discrepancy Optimization (AMP) to distinguish unknown class samples during online adaptation by amplifying the entropy difference between known and unknown samples. The study establishes a new benchmark derived from existing datasets with five modalities and evaluates AEO's performance across various domain shift scenarios, demonstrating its effectiveness in long-term and continual MM-OSTTA settings.
Researchers propose a spatio-temporal model for high-resolution wind forecasting in Saudi Arabia using Echo State Networks and stochastic partial differential equations. The model reduces spatial information via energy distance, captures dynamics with a sparse recurrent neural network, and reconstructs data using a non-stationary stochastic partial differential equation approach. The model achieves more accurate forecasts of wind speed and energy, potentially saving up to one million dollars annually compared to existing models.
This paper introduces Diffusion-BBO, a new online black-box optimization (BBO) framework that uses a conditional diffusion model as an inverse surrogate model. The framework employs an Uncertainty-aware Exploration (UaE) acquisition function to propose scores in the objective space for conditional sampling. The approach is shown theoretically to achieve a near-optimal solution and empirically outperforms existing online BBO baselines across 6 scientific discovery tasks.
Researchers at the Technology Innovation Institute (TII) have released a fully-annotated dataset for autonomous drone racing, called "Race Against the Machine." The dataset includes high-resolution visual, inertial, and motion capture data from both autonomous and piloted flights, along with commands, control inputs, and corner-level labeling of drone racing gates. The specifications to recreate their flight platform using commercial off-the-shelf components and the Betaflight controller are also released. Why it matters: This comprehensive resource aims to support the development of new methods and establish quantitative comparisons for approaches in robotics and AI, democratizing drone racing research.
The paper introduces a novel actor-critic framework called Distillation Policy Optimization that combines on-policy and off-policy data for reinforcement learning. It incorporates variance reduction mechanisms like a unified advantage estimator (UAE) and a residual baseline. The empirical results demonstrate improved sample efficiency for on-policy algorithms, bridging the gap with off-policy methods.
The paper proposes a method for causal inference using satellite image time series to determine the impact of interventions on climate change, focusing on quantifying deforestation due to human causes. The method uses computer vision and deep learning to detect forest tree coverage levels over time and Bayesian structural causal models to estimate counterfactuals. The framework is applied to analyze deforestation levels before and after the hyperinflation event in Brazil in the Amazon rainforest region.
KAUST and McLaren Racing have announced a five-year research partnership focused on R&D and extreme performance technology for Formula 1 cars. The collaboration will leverage KAUST's expertise in areas like sensors, electronics, numerical simulations, and fuel/engine combustion research. KAUST researchers will develop new experimental methods, mathematical models, and train students to understand complex systems. Why it matters: This partnership allows KAUST to apply its research to a real-world laboratory (Formula 1), fostering innovation in fuel technology, combustion, sensors, and algorithms with potential spillover effects for the broader automotive and engineering sectors in the region.
Researchers from MBZUAI, Khalifa University, and Sorbonne University Abu Dhabi developed H-SURF, a system of underwater robotic fish that can swim, communicate, and gather information without human guidance. The robotic fish use bioinspired robotics with streamlined bodies, fins, and propellers to produce fluid movement. They communicate with each other using light instead of sound to reduce noise. Why it matters: This award-winning system represents a significant advancement in autonomous underwater robotics, offering a less intrusive way to monitor marine environments and gather data, with potential applications in marine biology and environmental research.
MBZUAI researchers have developed SVRPBench, a new open benchmark for testing vehicle routing algorithms under real-world conditions. SVRPBench simulates unpredictable urban delivery scenarios including rush-hour traffic, accidents, and customer delivery time preferences. The benchmark uses realistic city models with clustered customer locations, unlike existing deterministic benchmarks. Why it matters: This benchmark offers a more practical evaluation for vehicle routing algorithms, potentially leading to significant cost savings and improved efficiency in logistics within the region and beyond.
MBZUAI researchers introduced CausalVerse, a new benchmark for causal representation learning (CRL) presented at NeurIPS 2025. CausalVerse combines high-fidelity visual complexity with access to underlying causal variables and graphs, featuring 200,000 images and 300 million video frames across 24 sub-scenes in four domains. It aims to provide a realistic and precise testbed to evaluate whether CRL methods can truly learn the right causes. Why it matters: By bridging the gap between toy datasets and real-world data, CausalVerse can drive advances in AI systems capable of understanding causality in complex scenarios.
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 Professor Sami Haddadin and his team developed a new framework called Tactile Skills to teach robots manual skills through touch and trial and error. This framework aims to address the gap in robots' ability to learn basic physical tasks compared to AI's advancements in language and image generation. The research, published in Nature Machine Intelligence, focuses on enabling robots to perform manipulation skills at industrial levels with low energy and compute demands. Why it matters: This research could lead to robots capable of performing household maintenance, industrial tasks, and even assisting in medical or rehabilitation settings, potentially solving labor shortages in various sectors in the region and beyond.
MBZUAI researchers have developed "Tactile Skills," a new embodied AI framework enabling robots to rapidly learn complex tactile tasks. The framework combines expert process knowledge with reusable tactile control and adaptation components, reducing reliance on extensive datasets. Tested on 28 industrial tasks, the robots achieved nearly 100% success, demonstrating adaptability to changing conditions. Why it matters: This breakthrough offers a practical and scalable approach to robotic automation, potentially transforming robots into adaptable assistants across diverse industries in the GCC.
MBZUAI is previewing PAN, a next-generation world model designed to simulate diverse realities and advance machine reasoning. PAN allows researchers to test AI agents in simulated environments before real-world deployment, enabling them to learn from mistakes without real-world consequences. It facilitates complex reasoning about actions, outcomes, and interactions, crucial for reliable AI performance in dynamic environments. Why it matters: PAN represents a significant advancement in AI by enabling comprehensive simulation and testing of AI agents, which can revolutionize fields like disaster management and healthcare where real-world experimentation is risky.
MBZUAI faculty, researchers, and students will present 34 papers at the 35th IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2023). Fahad Khan is a co-author on 11 accepted papers, while Salman Khan and Shijian Lu have 10 and 9 papers, respectively. One paper focuses on person image synthesis via a denoising diffusion model, and another introduces PromptCAL for generalized novel category discovery. Why it matters: This large volume of acceptances at a top-tier conference highlights MBZUAI's growing prominence and research contributions in computer vision, with potential impact across various industries from online retail to autonomous driving.
Researchers in Abu Dhabi developed H-SURF, a swarm of bio-inspired robotic fish for underwater data collection. Funded by the Technology Innovation Institute (TII) and conducted at Khalifa University, H-SURF uses swarm intelligence and optical communication to minimize disturbance to marine life. The project was recently recognized with the Sheikh Hamdan bin Zayed Award for Environmental Research.
The TUM Autonomous Motorsport team developed algorithms and deployment strategies for the Abu Dhabi Autonomous Racing League (A2RL). Their software emulates human driving behavior, pushing vehicle handling and multi-vehicle interactions. The team's approach led to a victory in the A2RL challenge. Why it matters: Autonomous racing serves as a valuable research environment for advancing autonomous driving tech and improving road safety in the region and globally.
This paper introduces a minimalistic autonomous racing stack designed for high-speed time-trial racing, emphasizing rapid deployment and efficient system integration with minimal on-track testing. Validated on real speedways, the stack achieved a top speed of 206 km/h within just 11 hours of practice, covering 325 km. The system performance analysis includes tracking accuracy, vehicle dynamics, and safety considerations. Why it matters: This research offers insights for teams aiming to quickly develop and deploy autonomous racing stacks with limited track access, potentially accelerating innovation in autonomous vehicle technology within the A2RL and similar racing initiatives.
This paper introduces a Bayesian optimization method for estimating tire parameters and their uncertainty, addressing a gap in existing literature. The methodology uses Stochastic Variational Inference to estimate parameters and uncertainties, and it is validated against a Nelder-Mead algorithm. The approach is applied to real-world data from the Abu Dhabi Autonomous Racing League, revealing uncertainties in identifying curvature and shape parameters due to insufficient excitation. Why it matters: The research provides a practical tool for assessing tire model parameters in real-world conditions, with implications for autonomous racing and vehicle dynamics modeling in the GCC region.
Researchers from the BME Formula Racing Team present the autonomous software stack of the FRED-003C, which enabled full-scale autonomous racing. The software stack was developed in the context of Formula Student Driverless competitions. The paper details the software pipeline, hardware-software architecture, and methods for perception, localization, mapping, planning, and control. Why it matters: The team's experience contributed to their participation in the Abu Dhabi Autonomous Racing League, and sharing the system provides a valuable starting point for other students in the region.
This paper introduces a longitudinal control system for autonomous racing vehicles with combustion engines, translating trajectory-tracking commands into low-level vehicle controls like throttle, brake pressure, and gear selection. The modular design facilitates integration with various trajectory-tracking algorithms and vehicles. Experimental validation on the EAV24 racecar during the Abu Dhabi Autonomous Racing League at Yas Marina Circuit demonstrated the system's effectiveness, achieving longitudinal accelerations up to 25 m/s². Why it matters: This research contributes to the advancement of autonomous racing technology in the region, showcasing practical applications in high-performance scenarios and fostering innovation in vehicle control systems.
Researchers introduce CESAR, a convolutional echo state autoencoder for high-resolution wind forecasting. The model extracts spatial features using a deep convolutional autoencoder and models their dynamics with an echo state network. Tested on high-resolution simulations in Riyadh, Saudi Arabia, CESAR improved wind speed and power forecasting by up to 17% compared to other methods. Why it matters: Accurate wind forecasting is critical for efficient wind farm planning and management in Saudi Arabia and the broader region.
Team TII EuroRacing (TII-ER) developed a full autonomous software stack for oval racing, enabling speeds above 75 m/s (270 km/h). The software includes modules for perception, planning, control, vehicle dynamics modeling, simulation, telemetry, and safety. The team achieved second and third place in the first two Indy Autonomous Challenge events using this stack.
Researchers at KAUST have developed a new method called Deep State Identifier for extracting information from videos for reinforcement learning. The method learns to predict returns from video-encoded episodes and identifies critical states using mask-based sensitivity analysis. Experiments demonstrate the method's potential for understanding and improving agent behavior in DRL.
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.
Team NimbRo presented four UAVs tailored for the MBZIRC 2020 challenges, including target chasing, wall building, and fire fighting. The UAVs utilized onboard object detection, aerial manipulation, LiDAR, and thermal cameras to perform their tasks autonomously. The team's software stack, which is mostly open-source, includes tools for system configuration, monitoring, and agile trajectory generation. Why it matters: The work demonstrates advanced robotics capabilities developed in the context of a major regional competition, advancing machine vision and trajectory generation, and showcasing potential applications in various sectors.
This paper presents the design and deployment of an autonomous unmanned ground vehicle (UGV) equipped with a robotic arm for urban firefighting. The UGV uses on-board sensors for navigation and a thermal camera for fire source identification, with a custom pump for fire suppression. The system was developed for the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2020, where it achieved the highest score among UGV solutions and contributed to winning first place. Why it matters: This demonstrates the potential of autonomous robotics in addressing complex and dangerous real-world challenges like urban firefighting in the GCC region and beyond.
A Carnegie Mellon team (Tartan) presented their approach to rapidly deployable and robust autonomous aerial vehicles at the 2020 Mohamed Bin Zayed International Robotics Challenge (MBZIRC). The system utilizes common techniques in vision and control, encoding robustness into mission structure through outcome monitoring and recovery strategies. Their system placed fourth in Challenge 2 and seventh in the Grand Challenge, with achievements in balloon popping, block manipulation, and autonomous firefighting. Why it matters: The work highlights strategies for building robust autonomous systems that can operate without central communication or high-precision GPS in challenging real-world environments, directly addressing key needs in the development of field robotics for the Middle East.
This paper presents a UAV-UGV team designed for autonomous firefighting, developed for the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2020. The system uses LiDAR for localization in GNSS-restricted environments and fuses LiDAR and thermal camera data to track fires. Relative navigation enables successful fire extinguishing. Why it matters: This research demonstrates the potential of robotic systems in autonomous firefighting, addressing challenges in dangerous and inaccessible environments, and advancing robotics research within the UAE.