MoCa: Modality-aware Continual Pre-training Makes Better Bidirectional Multimodal Embeddings

Open in new window