1 paper
Chenhao Zheng, Jieyu Zhang, Jianing Zhang +6
Tokenization in video models, typically through patchification, generates an excessive and redundant number of tokens. This severely limits video efficiency and scalability. While…