Fetching the paper…

Unified-IO 2: Scaling Autoregressive Multimodal Models with Vision, Language, Audio, and Action · Around