Kelix Technical Report
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917270583771136 |
|---|---|
| author | Ding, Boyang Chu, Chenglong Zang, Dunju Li, Han Cao, Jiangxia Gai, Kun Wei, Muhao Tang, Ruiming Wang, Shiyao Mao, Siyang Luo, Xinchen Liu, Yahui Ling, Zhixin Yang, Zhuoran Li, Ziming Song, Chengru Zhou, Guorui Zhang, Guowang Peng, Hao Wang, Hao Deng, Jiaxin Ouyang, Jin Zhang, Jinghao Ren, Lejian Wang, Qianqian Hu, Qigen Wang, Tao Wang, Xingmei Yang, Yiping Zhang, Zixing Wang, Ziqi |
| author_facet | Ding, Boyang Chu, Chenglong Zang, Dunju Li, Han Cao, Jiangxia Gai, Kun Wei, Muhao Tang, Ruiming Wang, Shiyao Mao, Siyang Luo, Xinchen Liu, Yahui Ling, Zhixin Yang, Zhuoran Li, Ziming Song, Chengru Zhou, Guorui Zhang, Guowang Peng, Hao Wang, Hao Deng, Jiaxin Ouyang, Jin Zhang, Jinghao Ren, Lejian Wang, Qianqian Hu, Qigen Wang, Tao Wang, Xingmei Yang, Yiping Zhang, Zixing Wang, Ziqi |
| contents | Autoregressive large language models (LLMs) scale well by expressing diverse tasks as sequences of discrete natural-language tokens and training with next-token prediction, which unifies comprehension and generation under self-supervision. Extending this paradigm to multimodal data requires a shared, discrete representation across modalities. However, most vision-language models (VLMs) still rely on a hybrid interface: discrete text tokens paired with continuous Vision Transformer (ViT) features. Because supervision is largely text-driven, these models are often biased toward understanding and cannot fully leverage large-scale self-supervised learning on non-text data. Recent work has explored discrete visual tokenization to enable fully autoregressive multimodal modeling, showing promising progress toward unified understanding and generation. Yet existing discrete vision tokens frequently lose information due to limited code capacity, resulting in noticeably weaker understanding than continuous-feature VLMs. We present Kelix, a fully discrete autoregressive unified model that closes the understanding gap between discrete and continuous visual representations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_09843 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Kelix Technical Report Ding, Boyang Chu, Chenglong Zang, Dunju Li, Han Cao, Jiangxia Gai, Kun Wei, Muhao Tang, Ruiming Wang, Shiyao Mao, Siyang Luo, Xinchen Liu, Yahui Ling, Zhixin Yang, Zhuoran Li, Ziming Song, Chengru Zhou, Guorui Zhang, Guowang Peng, Hao Wang, Hao Deng, Jiaxin Ouyang, Jin Zhang, Jinghao Ren, Lejian Wang, Qianqian Hu, Qigen Wang, Tao Wang, Xingmei Yang, Yiping Zhang, Zixing Wang, Ziqi Computer Vision and Pattern Recognition Autoregressive large language models (LLMs) scale well by expressing diverse tasks as sequences of discrete natural-language tokens and training with next-token prediction, which unifies comprehension and generation under self-supervision. Extending this paradigm to multimodal data requires a shared, discrete representation across modalities. However, most vision-language models (VLMs) still rely on a hybrid interface: discrete text tokens paired with continuous Vision Transformer (ViT) features. Because supervision is largely text-driven, these models are often biased toward understanding and cannot fully leverage large-scale self-supervised learning on non-text data. Recent work has explored discrete visual tokenization to enable fully autoregressive multimodal modeling, showing promising progress toward unified understanding and generation. Yet existing discrete vision tokens frequently lose information due to limited code capacity, resulting in noticeably weaker understanding than continuous-feature VLMs. We present Kelix, a fully discrete autoregressive unified model that closes the understanding gap between discrete and continuous visual representations. |
| title | Kelix Technical Report |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.09843 |