CoMa-3B

This repository contains CoMa-3B, a multimodal embedding model presented in the paper Compressing then Matching: An Efficient Pre-training Paradigm for Multimodal Embedding.

Introduction

CoMa (Compression then Matching) is an efficient pre-training paradigm designed to transform Multimodal Large Language Models (MLLMs) into competitive embedding models. It introduces a compressed pre-training phase that serves as a warm-up stage for contrastive learning, allowing the model to comprehensively preserve semantic content while emphasizing discriminative features for downstream tasks. CoMa achieves state-of-the-art results among MLLMs of comparable size on the MMEB benchmark.

• Paper: Compressing then Matching: An Efficient Pre-training Paradigm for Multimodal Embedding
• Repository: Trustworthy-Information-Access/CoMa

Architecture

The following diagram illustrates the CoMa architecture:

Citation

If you find this work useful, please consider citing:

@article{li2025compressing,
  title={Compressing then Matching: An Efficient Pre-training Paradigm for Multimodal Embedding},
  author={Li, Da and Luo, Yuxiao and Bi, Keping and Guo, Jiafeng and Yuan, Wei and Yang, Biao and Wang, Yan and Yang, Fan and Gao, Tingting and Zhou, Guorui},
  journal={arXiv preprint arXiv:2511.08480},
  year={2025}
}