Multimodal Long Video Modeling Based on Temporal Dynamic Context | Read Paper on Bytez