Paper Detail
Zhiheng Liu, Weiming Ren, Xiaoke Huang, Shoufa Chen, Tianhong Li, Mengzhao Chen, Yatai Ji, Sen He, Jonas Schult, Belinda Zeng, Tao Xiang, Wenhu Chen, Ping Luo, Luke Zettlemoyer, Yuren Cong
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.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@article{liu2026tuna,
title = {Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation},
author = {Zhiheng Liu and Weiming Ren and Xiaoke Huang and Shoufa Chen and Tianhong Li and Mengzhao Chen and Yatai Ji and Sen He and Jonas Schult and Belinda Zeng and Tao Xiang and Wenhu Chen and Ping Luo and Luke Zettlemoyer and Yuren Cong},
year = {2026},
abstract = {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 vi},
url = {https://arxiv.org/abs/2604.24763},
keywords = {cs.CV, multimodal models, vision encoders, visual representations, pixel embeddings, patch embedding layers, VAE, representation encoder, end-to-end optimization, latent-space approaches, pixel-space modelling, visual perception, huggingface daily},
eprint = {2604.24763},
archiveprefix = {arXiv},
}
{}