AVA-Encoder: Towards Agent-Native Video Representation Learning
By Chuyue Li · Paper · cs.CV
Creative agents still lack an effective way to learn from high-quality human films, limiting their ability to produce cinematic-grade videos. A key challenge is the absence of a structured video representation that is both faithful to film content and directly usable for agentic