Model
1 entry · 9 May 2023
Rather than needing paired training data across every modality combination, the model used images as a bridge to align the other five to a shared space.
Open weights & ecosystem · Models & capabilities