Cosmos 3 models are designed to process and generate language, images, video, audio and actions
A technical report on arXiv introduces Cosmos 3 as a family of world models for physical AI. The post sharing it discloses NVIDIA sponsorship.
TLDR
The Cosmos 3 technical report describes a family of world models designed to jointly process and generate language, image, video, audio and action sequences. The post sharing the report also links to NVIDIA's Cosmos and Cosmos Lab resources and states that it is sponsored by NVIDIA.
Combined views
291
1 Source, first seen 15d ago
saves