AI RESEARCH
Rethinking Position Embedding as a Context Controller for Multi-Reference and Multi-Shot Video Generation
arXiv CS.CV
•
ArXi:2604.03738v1 Announce Type: new