Welcome to the LCO-Embedding project - Scaling Language-centric Omnimodal Representation Learning.

Highlights:

• We introduce LCO-Embedding, a language-centric omnimodal representation learning method and the LCO-Embedding model families, setting a new state-of-the-art on MIEB (Massive Image Embedding Benchmark) while supporting audio and videos.
• We introduce the Generation-Representation Scaling Law, and connect models' generative capabilities and their representation upper bound.
• We introduce SeaDoc, a challenging visual document retrieval task in Southeast Asian languages, and show that continual generative pretraining before contrastive learning raises the representation upper bound.