MMEB-V3: Measuring the Performance Gaps of Omni-Modality Embedding Models

Haohang Huang, Xuan Lu, Mingyi Su +9

Multimodal embedding models aim to map heterogeneous inputs, such as text, images, videos, and audio, into a shared semantic space. However, existing methods and benchmarks remain…