Sparse-to-Dense: A Free Lunch for Lossless Acceleration of Video Understanding in LLMs

Open in new window