TUNA: Taming Unified Visual Representations for Native Unified Multimodal Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908684580290560 |
|---|---|
| author | Liu, Zhiheng Ren, Weiming Liu, Haozhe Zhou, Zijian Chen, Shoufa Qiu, Haonan Huang, Xiaoke An, Zhaochong Yang, Fanny Patel, Aditya Atliha, Viktar Ng, Tony Han, Xiao Zhu, Chuyan Zhang, Chenyang Liu, Ding Perez-Rua, Juan-Manuel He, Sen Schmidhuber, Jürgen Chen, Wenhu Luo, Ping Liu, Wei Xiang, Tao Schult, Jonas Cong, Yuren |
| author_facet | Liu, Zhiheng Ren, Weiming Liu, Haozhe Zhou, Zijian Chen, Shoufa Qiu, Haonan Huang, Xiaoke An, Zhaochong Yang, Fanny Patel, Aditya Atliha, Viktar Ng, Tony Han, Xiao Zhu, Chuyan Zhang, Chenyang Liu, Ding Perez-Rua, Juan-Manuel He, Sen Schmidhuber, Jürgen Chen, Wenhu Luo, Ping Liu, Wei Xiang, Tao Schult, Jonas Cong, Yuren |
| contents | Unified multimodal models (UMMs) aim to jointly perform multimodal understanding and generation within a single framework. We present TUNA, a native UMM that builds a unified continuous visual representation by cascading a VAE encoder with a representation encoder. This unified representation space allows end-to-end processing of images and videos for both understanding and generation tasks. Compared to prior UMMs with decoupled representations, TUNA's unified visual space avoids representation format mismatches introduced by separate encoders, outperforming decoupled alternatives in both understanding and generation. Moreover, we observe that stronger pretrained representation encoders consistently yield better performance across all multimodal tasks, highlighting the importance of the representation encoder. Finally, in this unified setting, jointly training on both understanding and generation data allows the two tasks to benefit from each other rather than interfere. Our extensive experiments on multimodal understanding and generation benchmarks show that TUNA achieves state-of-the-art results in image and video understanding, image and video generation, and image editing, demonstrating the effectiveness and scalability of its unified representation design. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_02014 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | TUNA: Taming Unified Visual Representations for Native Unified Multimodal Models Liu, Zhiheng Ren, Weiming Liu, Haozhe Zhou, Zijian Chen, Shoufa Qiu, Haonan Huang, Xiaoke An, Zhaochong Yang, Fanny Patel, Aditya Atliha, Viktar Ng, Tony Han, Xiao Zhu, Chuyan Zhang, Chenyang Liu, Ding Perez-Rua, Juan-Manuel He, Sen Schmidhuber, Jürgen Chen, Wenhu Luo, Ping Liu, Wei Xiang, Tao Schult, Jonas Cong, Yuren Computer Vision and Pattern Recognition Unified multimodal models (UMMs) aim to jointly perform multimodal understanding and generation within a single framework. We present TUNA, a native UMM that builds a unified continuous visual representation by cascading a VAE encoder with a representation encoder. This unified representation space allows end-to-end processing of images and videos for both understanding and generation tasks. Compared to prior UMMs with decoupled representations, TUNA's unified visual space avoids representation format mismatches introduced by separate encoders, outperforming decoupled alternatives in both understanding and generation. Moreover, we observe that stronger pretrained representation encoders consistently yield better performance across all multimodal tasks, highlighting the importance of the representation encoder. Finally, in this unified setting, jointly training on both understanding and generation data allows the two tasks to benefit from each other rather than interfere. Our extensive experiments on multimodal understanding and generation benchmarks show that TUNA achieves state-of-the-art results in image and video understanding, image and video generation, and image editing, demonstrating the effectiveness and scalability of its unified representation design. |
| title | TUNA: Taming Unified Visual Representations for Native Unified Multimodal Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2512.02014 |