Abstract: The advancement of Multimodal Large Language Models (MLLMs) has enabled significant progress in multi-modal understanding, expanding their capacity to analyze video content. However, ...