AI RESEARCH
Context Unrolling in Omni Models
arXiv CS.CV
•
ArXi:2604.21921v1 Announce Type: new We present Omni, a unified multimodal model natively trained on diverse modalities, including text, images, videos, 3D geometry, and hidden representations. We find that such