Borrowing treasures from neighbors: In-context learning for multimodal learning with missing modalities and data scarcity

Zhuo Zhi, Ziquan Liu, Moe Elbadawi, Adam Daneshmend, Mine Orlu, Abdul Basit, Andreas Demosthenous, Miguel R. D. Rodrigues

Abstract: Highlights•Address data scarcity in missing-modality tasks via adaptive data usage.•Propose data-dependent in-context learning to boost sample efficiency.•Achieve up to 10.8%<math><mtext is="true">%</mtext></math> gain over baselines on four datasets in low-data regime.