Despite significant advances in Multimodal Large Language Models (MLLMs), understanding complex temporal dynamics in videos remains a major challenge. Our experiments show that current Video Large Language Model (Video-LLM) architectures have critical limitations in temporal understanding, struggling with tasks that require detailed comprehension of action sequences and temporal progression. In this work, we propose a Video-LLM architecture that introduces stacked temporal attention modules directly within the vision encoder. This design incorporates a temporal attention in vision encoder, enabling the model to better capture the progression of actions and the relationships between frames before passing visual tokens to the LLM. Our results show that this approach significantly improves temporal reasoning and outperforms existing models in video question answering tasks, specifically in action recognition. We improve on benchmarks including VITATECS, MVBench, and Video-MME by up to +5.5%. By enhancing the vision encoder with temporal structure, we address a critical gap in video understanding for Video-LLMs. Project page and code are available at: https://alirasekh.github.io/STAVEQ2/

Energy efficiency is a critical design constraint in multicore Cyber-Physical Systems (CPS). Using energy management methods can violate timing constraints; hence, designing scheduling policies that can adapt to the dynamic and unpredictable nature of aperiodic real-time tasks remains a significant challenge. This paper introduces LEETMIC, a novel deep reinforcement learning framework for energy-aware real-time scheduling. To manage a variable number of active jobs, LEETMIC utilizes a learned policy network to determine the scheduling priority and Dynamic Voltage and Frequency Scaling (DVFS) level for each task individually, based on a combination of the task's local attributes and a summary of the global system state. The policy is trained using Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm to co-optimize for both task schedulability and energy consumption. The trained policy neural network is a compact multi-layer perceptron (MLP), making it suitable for online deployment in embedded systems. Experimental results demonstrate that, compared to the Global Earliest Deadline First (GEDF) scheduler, LEETMIC achieves a similar success ratio while significantly reducing energy consumption by 41.5% on average (up to 70%).