The field of multimodal learning is witnessing significant advancements, with a focus on improving the robustness and effectiveness of models in handling diverse multimedia data. Researchers are exploring innovative methods to address the challenges of modality bias, intra-modal feature extraction, and catastrophic forgetting in unified multimodal models. Novel frameworks and architectures are being proposed to enhance cross-modal understanding, mitigate forgetting, and improve performance in various tasks such as multimodal keyphrase generation, multimedia event extraction, and human action recognition. Notable papers in this area include: Augmenting Intra-Modal Understanding in MLLMs for Robust Multimodal Keyphrase Generation, which proposes a novel framework to reinforce intra-modal semantic learning in MLLMs. Stepwise Schema-Guided Prompting Framework with Parameter Efficient Instruction Tuning for Multimedia Event Extraction, which achieves state-of-the-art results in multimedia event extraction. Mitigating Intra- and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models, which introduces a lightweight and scalable architecture to mitigate both intra- and inter-modal forgetting. Heatmap Pooling Network for Action Recognition from RGB Videos, which proposes a novel heatmap pooling network for action recognition from videos. Detection of Intoxicated Individuals from Facial Video Sequences via a Recurrent Fusion Model, which introduces a novel video-based facial sequence analysis approach for detecting alcohol intoxication. Towards Adaptive Fusion of Multimodal Deep Networks for Human Action Recognition, which explores adaptive fusion strategies across multiple modalities for human action recognition.