LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs - Explained Simply

LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs - Explained Simply | ArXiv Explained