Wings: Learning Multimodal LLMs without Text-only Forgetting

Open in new window