(논문 요약) AnyMAL: An Efficient and Scalable Any-Modality Augmented Language Model (Paper)

핵심 내용

• 다양한 modality 를 LLM 과 연동함.
• a lightweight adapter for each modality (projection layer)

실험 결과