Latent Communication between Language Models: Multimodal Communication

Can two language models trained on different modalities (e.g., image and text) communicate in latent space to unlock better visual reasoning?

Mootez Saad

Project lead · Dalhousie University