Kwai Keye-VL 1.5 Technical Report
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_ | 1866918137234980864 |
|---|---|
| author | Yang, Biao Wen, Bin Ding, Boyang Liu, Changyi Chu, Chenglong Song, Chengru Rao, Chongling Yi, Chuan Li, Da Zang, Dunju Yang, Fan Zhou, Guorui Zhang, Guowang Shen, Han Peng, Hao Ding, Haojie Wang, Hao Fan, Haonan Ju, Hengrui Huang, Jiaming Cao, Jiangxia Chen, Jiankang Hua, Jingyun Chen, Kaibing Jiang, Kaiyu Tang, Kaiyu Gai, Kun Wei, Muhao Wang, Qiang Wang, Ruitao Na, Sen Zhang, Shengnan Mao, Siyang Huang, Sui Zhang, Tianke Gao, Tingting Chen, Wei Yuan, Wei Wu, Xiangyu Hu, Xiao Lu, Xingyu Zhang, Yi-Fan Yang, Yiping Chen, Yulong Lu, Zeyi Wu, Zhenhua Ling, Zhixin Yang, Zhuoran Li, Ziming Xu, Di Gao, Haixuan Li, Hang Wang, Jing Ren, Lejian Hu, Qigen Wang, Qianqian Wang, Shiyao Luo, Xinchen Li, Yan Hu, Yuhang Zhang, Zixing |
| author_facet | Yang, Biao Wen, Bin Ding, Boyang Liu, Changyi Chu, Chenglong Song, Chengru Rao, Chongling Yi, Chuan Li, Da Zang, Dunju Yang, Fan Zhou, Guorui Zhang, Guowang Shen, Han Peng, Hao Ding, Haojie Wang, Hao Fan, Haonan Ju, Hengrui Huang, Jiaming Cao, Jiangxia Chen, Jiankang Hua, Jingyun Chen, Kaibing Jiang, Kaiyu Tang, Kaiyu Gai, Kun Wei, Muhao Wang, Qiang Wang, Ruitao Na, Sen Zhang, Shengnan Mao, Siyang Huang, Sui Zhang, Tianke Gao, Tingting Chen, Wei Yuan, Wei Wu, Xiangyu Hu, Xiao Lu, Xingyu Zhang, Yi-Fan Yang, Yiping Chen, Yulong Lu, Zeyi Wu, Zhenhua Ling, Zhixin Yang, Zhuoran Li, Ziming Xu, Di Gao, Haixuan Li, Hang Wang, Jing Ren, Lejian Hu, Qigen Wang, Qianqian Wang, Shiyao Luo, Xinchen Li, Yan Hu, Yuhang Zhang, Zixing |
| contents | In recent years, the development of Large Language Models (LLMs) has significantly advanced, extending their capabilities to multimodal tasks through Multimodal Large Language Models (MLLMs). However, video understanding remains a challenging area due to the dynamic and information-dense nature of videos. Existing models struggle with the trade-off between spatial resolution and temporal coverage when processing video content. We present Keye-VL-1.5, which addresses fundamental challenges in video comprehension through three key innovations. First, we introduce a novel Slow-Fast video encoding strategy that dynamically allocates computational resources based on inter-frame similarity, processing key frames with significant visual changes at higher resolution (Slow pathway) while handling relatively static frames with increased temporal coverage at lower resolution (Fast pathway). Second, we implement a progressive four-stage pre-training methodology that systematically extends the model's context length from 8K to 128K tokens, enabling processing of longer videos and more complex visual content. Third, we develop a comprehensive post-training pipeline focusing on reasoning enhancement and human preference alignment, incorporating a 5-step chain-of-thought data construction process, iterative GSPO-based reinforcement learning with progressive prompt hinting for difficult cases, and alignment training. Through extensive evaluation on public benchmarks and rigorous internal human assessment, Keye-VL-1.5 demonstrates significant improvements over existing models, particularly excelling in video understanding tasks while maintaining competitive performance on general multimodal benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_01563 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Kwai Keye-VL 1.5 Technical Report Yang, Biao Wen, Bin Ding, Boyang Liu, Changyi Chu, Chenglong Song, Chengru Rao, Chongling Yi, Chuan Li, Da Zang, Dunju Yang, Fan Zhou, Guorui Zhang, Guowang Shen, Han Peng, Hao Ding, Haojie Wang, Hao Fan, Haonan Ju, Hengrui Huang, Jiaming Cao, Jiangxia Chen, Jiankang Hua, Jingyun Chen, Kaibing Jiang, Kaiyu Tang, Kaiyu Gai, Kun Wei, Muhao Wang, Qiang Wang, Ruitao Na, Sen Zhang, Shengnan Mao, Siyang Huang, Sui Zhang, Tianke Gao, Tingting Chen, Wei Yuan, Wei Wu, Xiangyu Hu, Xiao Lu, Xingyu Zhang, Yi-Fan Yang, Yiping Chen, Yulong Lu, Zeyi Wu, Zhenhua Ling, Zhixin Yang, Zhuoran Li, Ziming Xu, Di Gao, Haixuan Li, Hang Wang, Jing Ren, Lejian Hu, Qigen Wang, Qianqian Wang, Shiyao Luo, Xinchen Li, Yan Hu, Yuhang Zhang, Zixing Computer Vision and Pattern Recognition In recent years, the development of Large Language Models (LLMs) has significantly advanced, extending their capabilities to multimodal tasks through Multimodal Large Language Models (MLLMs). However, video understanding remains a challenging area due to the dynamic and information-dense nature of videos. Existing models struggle with the trade-off between spatial resolution and temporal coverage when processing video content. We present Keye-VL-1.5, which addresses fundamental challenges in video comprehension through three key innovations. First, we introduce a novel Slow-Fast video encoding strategy that dynamically allocates computational resources based on inter-frame similarity, processing key frames with significant visual changes at higher resolution (Slow pathway) while handling relatively static frames with increased temporal coverage at lower resolution (Fast pathway). Second, we implement a progressive four-stage pre-training methodology that systematically extends the model's context length from 8K to 128K tokens, enabling processing of longer videos and more complex visual content. Third, we develop a comprehensive post-training pipeline focusing on reasoning enhancement and human preference alignment, incorporating a 5-step chain-of-thought data construction process, iterative GSPO-based reinforcement learning with progressive prompt hinting for difficult cases, and alignment training. Through extensive evaluation on public benchmarks and rigorous internal human assessment, Keye-VL-1.5 demonstrates significant improvements over existing models, particularly excelling in video understanding tasks while maintaining competitive performance on general multimodal benchmarks. |
| title | Kwai Keye-VL 1.5 Technical Report |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.01563 |