Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916019802472448 |
|---|---|
| author | Liu, Zhiheng Ren, Weiming Huang, Xiaoke Chen, Shoufa Li, Tianhong Chen, Mengzhao Ji, Yatai He, Sen Schult, Jonas Zeng, Belinda Xiang, Tao Chen, Wenhu Luo, Ping Zettlemoyer, Luke Cong, Yuren |
| author_facet | Liu, Zhiheng Ren, Weiming Huang, Xiaoke Chen, Shoufa Li, Tianhong Chen, Mengzhao Ji, Yatai He, Sen Schult, Jonas Zeng, Belinda Xiang, Tao Chen, Wenhu Luo, Ping Zettlemoyer, Luke Cong, Yuren |
| contents | Unified multimodal models typically rely on pretrained vision encoders and use separate visual representations for understanding and generation, creating misalignment between the two tasks and preventing fully end-to-end optimization from raw pixels. We introduce Tuna-2, a native unified multimodal model that performs visual understanding and generation directly based on pixel embeddings. Tuna-2 drastically simplifies the model architecture by employing simple patch embedding layers to encode visual input, completely discarding the modular vision encoder designs such as the VAE or the representation encoder. Experiments show that Tuna-2 achieves state-of-the-art performance in multimodal benchmarks, demonstrating that unified pixel-space modelling can fully compete with latent-space approaches for high-quality image generation. Moreover, while the encoder-based variant converges faster in early pretraining, Tuna-2's encoder-free design achieves stronger multimodal understanding at scale, particularly on tasks requiring fine-grained visual perception. These results show that pretrained vision encoders are not necessary for multimodal modelling, and end-to-end pixel-space learning offers a scalable path toward stronger visual representations for both generation and perception. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_24763 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation Liu, Zhiheng Ren, Weiming Huang, Xiaoke Chen, Shoufa Li, Tianhong Chen, Mengzhao Ji, Yatai He, Sen Schult, Jonas Zeng, Belinda Xiang, Tao Chen, Wenhu Luo, Ping Zettlemoyer, Luke Cong, Yuren Computer Vision and Pattern Recognition Unified multimodal models typically rely on pretrained vision encoders and use separate visual representations for understanding and generation, creating misalignment between the two tasks and preventing fully end-to-end optimization from raw pixels. We introduce Tuna-2, a native unified multimodal model that performs visual understanding and generation directly based on pixel embeddings. Tuna-2 drastically simplifies the model architecture by employing simple patch embedding layers to encode visual input, completely discarding the modular vision encoder designs such as the VAE or the representation encoder. Experiments show that Tuna-2 achieves state-of-the-art performance in multimodal benchmarks, demonstrating that unified pixel-space modelling can fully compete with latent-space approaches for high-quality image generation. Moreover, while the encoder-based variant converges faster in early pretraining, Tuna-2's encoder-free design achieves stronger multimodal understanding at scale, particularly on tasks requiring fine-grained visual perception. These results show that pretrained vision encoders are not necessary for multimodal modelling, and end-to-end pixel-space learning offers a scalable path toward stronger visual representations for both generation and perception. |
| title | Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.24763 |