Nomic Embed Vision: Expanding the Latent Space

TL;DR

This paper introduces nomic-embed-vision, an image embedding model that shares a unified latent space with text embeddings, enabling high performance across vision, language, and multimodal tasks.

Contribution

It presents the first unified latent space for vision and language embeddings, combining image and text models for improved multimodal performance.

Findings

Shared latent space enhances multimodal task performance

Open-source model facilitates research and application development

Achieves high accuracy in vision and language tasks

Abstract

This technical report describes the training of nomic-embed-vision, a highly performant, open-code, open-weights image embedding model that shares the same latent space as nomic-embed-text. Together, nomic-embed-vision and nomic-embed-text form the first unified latent space to achieve high performance across vision, language, and multimodal tasks.