[KongchangAI]
Concept

Multimodal Retrieval

A retrieval technology that simultaneously processes multiple information modalities such as text, images, video and speech, requiring cross-modal matching across multiple libraries.

Timeline (last 90 days)

Oct 26

The method reduces the ℓ2 distance between embedding centroids by 97–99% on five multimodal retrieval benchmarks.

Unverified50%

All Facts (1)

Source Articles