Tian-Le Liu

ELASTIC-V: Reconstructing Spatiotemporal Fidelity via Saliency Compression and Elastic Encoding

While Multimodal Large Language Models (MLLMs) excel in short video understanding, extending them to long videos remains challenging due to memory bottlenecks and temporal hallucinations. Existing uniform sampling and discrete pruning methods often disrupt temporal boundaries and distort physical time scales. To addres...