Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2408.15998 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913713889476608 |
|---|---|
| author | Shi, Min Liu, Fuxiao Wang, Shihao Liao, Shijia Radhakrishnan, Subhashree Zhao, Yilin Huang, De-An Yin, Hongxu Sapra, Karan Yacoob, Yaser Shi, Humphrey Catanzaro, Bryan Tao, Andrew Kautz, Jan Yu, Zhiding Liu, Guilin |
| author_facet | Shi, Min Liu, Fuxiao Wang, Shihao Liao, Shijia Radhakrishnan, Subhashree Zhao, Yilin Huang, De-An Yin, Hongxu Sapra, Karan Yacoob, Yaser Shi, Humphrey Catanzaro, Bryan Tao, Andrew Kautz, Jan Yu, Zhiding Liu, Guilin |
| contents | The ability to accurately interpret complex visual information is a crucial topic of multimodal large language models (MLLMs). Recent work indicates that enhanced visual perception significantly reduces hallucinations and improves performance on resolution-sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_15998 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Eagle: Exploring The Design Space for Multimodal LLMs with Mixture of Encoders Shi, Min Liu, Fuxiao Wang, Shihao Liao, Shijia Radhakrishnan, Subhashree Zhao, Yilin Huang, De-An Yin, Hongxu Sapra, Karan Yacoob, Yaser Shi, Humphrey Catanzaro, Bryan Tao, Andrew Kautz, Jan Yu, Zhiding Liu, Guilin Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Robotics The ability to accurately interpret complex visual information is a crucial topic of multimodal large language models (MLLMs). Recent work indicates that enhanced visual perception significantly reduces hallucinations and improves performance on resolution-sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks. |
| title | Eagle: Exploring The Design Space for Multimodal LLMs with Mixture of Encoders |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Robotics |
| url | https://arxiv.org/abs/2408.15998 |