Multi-modal Generative Models

Multi-modal Perceiver

Unified architecture that processes different modalities (text, image, audio) as token sequences in a single latent space, using cross-attentions to model their interactions.

← 뒤로