UniME-V2: MLLM-as-a-Judge for Universal Multimodal Embedding Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911305899704320 |
|---|---|
| author | Gu, Tiancheng Yang, Kaicheng Zhang, Kaichen An, Xiang Feng, Ziyong Zhang, Yueyi Cai, Weidong Deng, Jiankang Bing, Lidong |
| author_facet | Gu, Tiancheng Yang, Kaicheng Zhang, Kaichen An, Xiang Feng, Ziyong Zhang, Yueyi Cai, Weidong Deng, Jiankang Bing, Lidong |
| contents | Universal multimodal embedding models are foundational to various tasks. Existing approaches typically employ in-batch negative mining by measuring the similarity of query-candidate pairs. However, these methods often struggle to capture subtle semantic differences among candidates and lack diversity in negative samples. Moreover, the embeddings exhibit limited discriminative ability in distinguishing false and hard negatives. In this paper, we leverage the advanced understanding capabilities of MLLMs to enhance representation learning and present a novel Universal Multimodal Embedding (UniME-V2) model. Our approach first constructs a potential hard negative set through global retrieval. We then introduce the MLLM-as-a-Judge mechanism, which utilizes MLLMs to assess the semantic alignment of query-candidate pairs and generate soft semantic matching scores. These scores serve as a foundation for hard negative mining, mitigating the impact of false negatives and enabling the identification of diverse, high-quality hard negatives. Furthermore, the semantic matching scores are used as soft labels to mitigate the rigid one-to-one mapping constraint. By aligning the similarity matrix with the soft semantic matching score matrix, the model learns semantic distinctions among candidates, significantly enhancing its discriminative capacity. To further improve performance, we propose UniME-V2-Reranker, a reranking model trained on our mined hard negatives through a joint pairwise and listwise optimization approach. We conduct comprehensive experiments on the MMEB benchmark and multiple retrieval tasks, demonstrating that our method achieves state-of-the-art performance on average across all tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_13515 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | UniME-V2: MLLM-as-a-Judge for Universal Multimodal Embedding Learning Gu, Tiancheng Yang, Kaicheng Zhang, Kaichen An, Xiang Feng, Ziyong Zhang, Yueyi Cai, Weidong Deng, Jiankang Bing, Lidong Computer Vision and Pattern Recognition Artificial Intelligence Universal multimodal embedding models are foundational to various tasks. Existing approaches typically employ in-batch negative mining by measuring the similarity of query-candidate pairs. However, these methods often struggle to capture subtle semantic differences among candidates and lack diversity in negative samples. Moreover, the embeddings exhibit limited discriminative ability in distinguishing false and hard negatives. In this paper, we leverage the advanced understanding capabilities of MLLMs to enhance representation learning and present a novel Universal Multimodal Embedding (UniME-V2) model. Our approach first constructs a potential hard negative set through global retrieval. We then introduce the MLLM-as-a-Judge mechanism, which utilizes MLLMs to assess the semantic alignment of query-candidate pairs and generate soft semantic matching scores. These scores serve as a foundation for hard negative mining, mitigating the impact of false negatives and enabling the identification of diverse, high-quality hard negatives. Furthermore, the semantic matching scores are used as soft labels to mitigate the rigid one-to-one mapping constraint. By aligning the similarity matrix with the soft semantic matching score matrix, the model learns semantic distinctions among candidates, significantly enhancing its discriminative capacity. To further improve performance, we propose UniME-V2-Reranker, a reranking model trained on our mined hard negatives through a joint pairwise and listwise optimization approach. We conduct comprehensive experiments on the MMEB benchmark and multiple retrieval tasks, demonstrating that our method achieves state-of-the-art performance on average across all tasks. |
| title | UniME-V2: MLLM-as-a-Judge for Universal Multimodal Embedding Learning |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2510.13515 |