UniVid: The Open-Source Unified Video Model
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912616799010816 |
|---|---|
| author | Luo, Jiabin Lin, Junhui Zhang, Zeyu Wu, Biao Fang, Meng Chen, Ling Tang, Hao |
| author_facet | Luo, Jiabin Lin, Junhui Zhang, Zeyu Wu, Biao Fang, Meng Chen, Ling Tang, Hao |
| contents | Unified video modeling that combines generation and understanding capabilities is increasingly important but faces two key challenges: maintaining semantic faithfulness during flow-based generation due to text-visual token imbalance and the limitations of uniform cross-modal attention across the flow trajectory, and efficiently extending image-centric MLLMs to video without costly retraining. We present UniVid, a unified architecture that couples an MLLM with a diffusion decoder through a lightweight adapter, enabling both video understanding and generation. We introduce Temperature Modality Alignment to improve prompt adherence and Pyramid Reflection for efficient temporal reasoning via dynamic keyframe selection. Extensive experiments on standard benchmarks demonstrate state-of-the-art performance, achieving a 2.2% improvement on VBench-Long total score compared to EasyAnimateV5.1, and 1.0% and 3.3% accuracy gains on MSVD-QA and ActivityNet-QA, respectively, compared with the best prior 7B baselines. Code: https://github.com/AIGeeksGroup/UniVid. Website: https://aigeeksgroup.github.io/UniVid. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_24200 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | UniVid: The Open-Source Unified Video Model Luo, Jiabin Lin, Junhui Zhang, Zeyu Wu, Biao Fang, Meng Chen, Ling Tang, Hao Computer Vision and Pattern Recognition Unified video modeling that combines generation and understanding capabilities is increasingly important but faces two key challenges: maintaining semantic faithfulness during flow-based generation due to text-visual token imbalance and the limitations of uniform cross-modal attention across the flow trajectory, and efficiently extending image-centric MLLMs to video without costly retraining. We present UniVid, a unified architecture that couples an MLLM with a diffusion decoder through a lightweight adapter, enabling both video understanding and generation. We introduce Temperature Modality Alignment to improve prompt adherence and Pyramid Reflection for efficient temporal reasoning via dynamic keyframe selection. Extensive experiments on standard benchmarks demonstrate state-of-the-art performance, achieving a 2.2% improvement on VBench-Long total score compared to EasyAnimateV5.1, and 1.0% and 3.3% accuracy gains on MSVD-QA and ActivityNet-QA, respectively, compared with the best prior 7B baselines. Code: https://github.com/AIGeeksGroup/UniVid. Website: https://aigeeksgroup.github.io/UniVid. |
| title | UniVid: The Open-Source Unified Video Model |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.24200 |