The field of video understanding and representation is experiencing significant advancements, driven by developments in action detection, video representation learning, and vision-language models. A common theme among these areas is the shift towards more open-vocabulary, view-invariant, and self-supervised approaches. Researchers are exploring ways to reduce the reliance on parameter-heavy architectures and large-scale datasets, instead focusing on efficient and adaptable models. Notable innovations include novel curriculum learning procedures, knowledge distillation objectives, and weakly-supervised and few-shot learning methods. The use of variational inference and self-supervised learning is also becoming increasingly popular, with applications in clustering, representation learning, and symmetry discovery. Furthermore, multimodal large language models are being improved with techniques such as reinforcement fine-tuning, probabilistic jump diffusion, and gaze consensus adaptation, enabling better reasoning and understanding of complex visual and temporal information. Overall, these developments have the potential to significantly advance the field of video understanding and representation, with applications in areas such as video question answering, activity recognition, and content retrieval. Key papers in these areas include Scaling Open-Vocabulary Action Detection, Learning Activity View-invariance Under Extreme Viewpoint Changes via Curriculum Knowledge Distillation, LV-MAE, AutoSSVH, VIP, FASR-Net, VideoAgent2, REEF, REVEAL, and LVC.